mit :: whirlwind :: Redmond Project Whirlwind A Case History In Contemporary Technology 1975

Ht-K+fiNH

.s«^

*>i<*®«

»■ i •

►J+^+^+l

4^ •* * ■

MH

l+B«HMB'

L4J

fe*&

iHf^ , IvX *"T /^ t>

<@*1

1*1
MM

ij^T 7

2 ^

•e*

PROJECT WHIRLWIND

A Case History in
Contemporary Technology

Kent C. Redmond and Thomas M. Smith

Reproduced by

The MITRE Corporation

Bedford, MA

November 1975

FOREWORD

In the beginning, MIT begat Whirlwind. Whirlwind begat SAGE;
SAGE begat Lincoln Laboratory; Lincoln Laboratory begat MITRE. Lest
our lineage be forgot, we publish the Whirlwind History.

The Whirlwind History was written in 1967 by Kent Redmond and
Tom Smith on a grant from The MITRE Corporation. It was intended for
publication by the Smithsonian Institution as part of a series on the history
of computer development, but when the idea of the series was dropped by
the Smithsonian, the manuscript lay fallow for a number of years. Fre-
quent requests for copies were honored by photocopy of photocopy, with
the result that legibility was poor and there were delays in production.

We managed to locate an original copy and have reproduced a few
copies in the interest of preserving this well done piece of the computer
story for future scholars and historians.

Robert R. Everett
President, The MITRE Corporation

m

TABLE OF CONTENTS

Chapter Page

1 The Beginning 1.1

2 Computing Problems Emerge 2.1

3 The Shift to Digital 3. 1

4 Preliminary Design Efforts 4. 1

5 Pressure from ONR 5.1

6 Problems of Federal Assistance 6. 1

7 Breaking New Trails 7. 1

8 R&D Policies and Practices 8. 1

9 The Collision Course of ONR and Whirlwind 9. 1

10 ADSEC and Whirlwind 10. 1

11 Internal Storage Problems 11.1

12 Magnetic Cores and R&D Progress 12. 1

13 In Retrospect 13.1

Chapter One
THE BEGINNING

Since the Second World War there has been growing
recognition in the United States of the practical value
of science for the defense and welfare of the American
people. Along with this appreciation of the usefulness
of applied science there has been also a growing appre-
hensiveness at the rapid and profound changes being
wrought in our society by the multibillion-dollar
scientific technology we have created. Government,
industry, and institutions of higher education have pooled
their dollar and manpower resources with great ingenuity
to provide awesome weapons of war and magnificent pro-
duction, transportation, and communication facilities,
and it has all happened so fast that neither the experts
nor the common citizens are always sure what we are
doing, where we are going, or which direction we should
be heading as a society when putting our scientific
technology to work for us .

The immediate impact of our scientific and technological
achievements has not been hard to imagine and anticipate ,
although sometimes the force of the impact and the rate
and scale of the direct consequences have surprised us.

1.01

1.02

Of greater import and harder to foresee have been the
second-order and third-order effects, the tumbling-domino
consequences, and the cumulative alterations that threaten
to move us into patterns of living we do not like, do
not understand, and do not want.

This historical study of one research and development
project takes a look at an example of what we are doing
and where we are going on a small scale, and to these ends
it examines a project of our scientific technology,
Project Whirlwind, in some detail in order to cast some
light — again on a small scale — in one direction we are
heading, that of the information revolution via the
research and development road. Whether this particular
project more fittingly provides information and examples
of how the business of modern scientific technology ought
to be conducted, or whether it offers instructive,
cautionary lessons in what should not be done is for the
reader to decide, although the authors offer their own
judgments and conclusions.

This story is written not for the technical specialist
or the management specialist or the funding specialist
but for the thoughtful layman who believes, as do the
authors, that if we gain a clearer understanding of what
we are doing and what we can do, this will help us decide
where we want to go and how to get there, in those affairs
that require the use of scientific technology as our
obedient servant.

1.03

The men in this story were engineers. Their aim in
1944 was to design and build an aircraft simulator, but
their achievement by 1953 was as different as it was
unforeseen. Instead, in the course of a decade of
pioneering electronic and radar experience , they had
acquired a thoroughgoing mastery of the basic concepts of
integrated system design; they had built "Whirlwind," a
high-speed, prototypal, digital computer that became
uniquely appropriate for a brief, mid-century strategic
mission in the defense of the nation; and they had created
a small group of experts who by their contributions then
and later were to seed American computer technology with
a know-how that , in the reckoning of some observers ,
transformed the computer overnight from a limited instrument
intended primarily for mathematical and scientific
computation to a device of wide and practical social
potentiality.

In accomplishing this preliminary transformation,
these men vaulted the technical computer state of the art
a decade ahead of where it otherwise would have been,
according to this view. Their contributions immensely
strengthened the sinews of the emerging computer technology
for the tasks that lay ahead. While it was true that
their actions unexpectedly accelerated the onset of the
unanticipated information revolution for which this century
appears likely to be remembered by later generations,

1.04

nevertheless the net gains in ability to identify, define,
and offer solutions to social problems could be expected
to more than compensate for the stresses and uncertainties
which that revolution and the general onrush of human
affairs would impose.

On the debit side, in the view of others, the men of
Project Whirlwind extravagantly spent some five million
dollars of public money in five short years. They pursued
impetuous, risky, and unrealistic research and development
practices in peace time, practices of the sort that
prudently can be sustained by a nation only on a short-
term basis* in time of war and extreme crisis. Their
project was able to flourish in special and unique
circumstances ,. like an experimental hothouse plant in
forced growth, , and such favorable circumstances are
quite unlike the conditions that normally prevail in the
conduct of research and development affairs .

Consequently, this project was not typical, not
representative, and not an exemplar to be followed. It
was at best a lesson in fortunate improvisation, and it
offered a clear warning to all how runaway tendencies can
dominate the enthusiastic pursuit of research and develop-
ment: when business-as-usual restraints are absent* According
to this view* Project Whirlwind provided not a lesson in
how the efficient and expeditious conduct of research and
development might be achieved as a new norm, but a

1.05

demonstration , by its obvious malpractices , of the
essential wisdom of traditional procedures.

It succeeded rather than failed, according to this
argument, because of unusual and unexpected circumstances
beyond its control. The project had become an engineering-
development project without a practical mission until
these circumstances, involving a potential shift in the
very balance of international power in world affairs,
had intervened. Not only was the project not a business-
as-usual enterprise, but it took nothing less than a
looming national military and political crisis to come
to its rescue. Had Project Whirlwind been conceived in
the beginning or shortly thereafter been modified in
anticipation of this crisis, then its importance, its
priority rank, and its conduct of its own affairs would
have developed naturally. Instead, one could argue, it
had been fiscally hell-bent to develop a fantastic machine
for which virtually no one except its enthusiastic builders
could see any use.

While pure scientists might be excused for spending
modest sums for their traditionally impractical investiga-
tions of the unknown, even though the ultimate practical
payoff was not visible, engineers setting up expensive
development and experimental-prototype projects must not
be allowed to proceed without an explicit, agreed-upon,
practical goal in view. Pure science could afford to
leap-frog ahead into the unknown, because if some of its

1.06

enterprises did fall on their faces, the national loss
in dollars and manhours would be minimal and tolerable.
Not so, where multimillion-dollar engineering projects
such as Whirlwind were concerned, and Whirlwind had just
missed, by perhaps a hair's breadth, falling on its face.
Even its magnificent, internal, magnetic-core storage had
emerged as a desperate, risky, ad hoc engineering solution
to the nagging problems of unreliable electrostatic-tube
storage. Project Whirlwind, when all was said and done,
had been lucky.

Or had it?

It is possible to take a position of praise or a
position of censure or one that is a mixture of the two.
But whatever the convictions of the observer and whatever
the verdict, this research and development project offers
a significant history for the thoughtful observer of the
research and davelopment process . The genesis and develop-
ment of the project were characterized by a mixture of
elements that were traditional and elements that were novel.
Thus, the men did not work in the relatively independent
entrepreneurial isolation that characterized, for example,
the creative efforts of the Wright brothers . Instead ,
they worked under an institutional aegis. Nominally the
aegis was that of higher education, for it was provided
by the Massachusetts Institute of Technology. But actually
it was a peculiar combination of educational and governmental,

1.07

tinged with industrial, and it was made possible only
by the unprecedented exigencies and modes of activity
created by the Second World War.

While MIT furnished a physical plant and the
technically trained intellectual resources, the United
States Navy during the earlier period and the United
States Air Force during the later period furnished the
necessary funds. National-defense tax dollars thus
entirely underwrote the cost of this enterprise in
twentieth-century scientific research and engineering
development.

That this should be the case was a natural conse-
quence of the historical circumstances that had impelled
the United States into the war. Well before 19 44,
prosecution of the war against Nazi Germany and Japan
had brought numerous technical problems and their solutions
to the attention of American engineers and scientists.
To an extent previously unknown, American engineers and
scientists were providing essential technical leadership
while cooperating with the national military establish-
ment and American industry. Through such agencies as
the National Defense Research Committee and the Office
of Scientific Research and Development, some 30,0 00
engineers and scientists became, in the words of OSRD
Director Vannevar Bush, "full and responsible partners
for the first time in the conduct of war." Approximately

1.08

one-half billion dollars was expended by these agencies
in the search for new weapons and new medicines .

The Japanese attack upon Pearl Harbor in December,
1941, caught both the armed forces and industry unprepared
for the vital responsibilities thrust so abruptly upon
them. Presidential efforts to make of the United States
the "arsenal of democracy," given substance in March,
1911 by the enactment of Lend-Lease legislation, had,
it is true, encouraged the efforts of American industry
to increase substantially the production of arms and
other military equipment for the beleaguered forces of
the United Kingdom, but such efforts were a trifle com-
pared to the needs created by the precipitation of the
United States into the conflict. American entry demanded
the immediate and total conversion of the national
industrial, technical, and scientific complex to the
development and manufacture of the weapons and ancillary
equipment which were to become the instruments of victory.
Since the nation's colleges and universities were the
repositories of much of its scientific and engineering
talent, the Government turned to them to obtain "large-
scale assistance . . . mainly for military applications
of nuclear energy, communications, control systems, and
improvements in propulsion." These institutions responded
to the call, adapting themselves to meet the vital
challenge and rendering such aid and leadership as they
could.

1.09

Among these institutions was the Massachusetts
Institute of Technology, recognized as a leading edu-
cational and research center in science and engineering.
Even prior to American entry in 1941 MIT, as part of
its contribution to the war against fascism, had enlarged
its areas of scientific research and engineering develop-
ment by the addition of programs and facilities designed
specifically to seek solutions to technical problems
arising from the need for new and improved weapons. One

of the facilities added was the Servomechanisms Laboratory
of the Department of Electrical Engineering. The special
competence developed by the Laboratory and its personnel,
coupled with the technical resources of the Institute, had
formed by 1944 a combination of talent uniquely qualified to
undertake for the United States Navy a project that ultimately
was to make a major contribution to computer technology.

As originally conceived, the project would provide
a common solution to a twofold problem, that of the
flight instructor and that of the aircraft designer.
It was taking far too much time and money to train
flight crews to man the more complex, newer models, and
it was taking far too much time and money to design
projected high-performance airplanes. A possible solution
worth investigating had been suggested by the recent
successful development of flight trainers.

The massive trained-manpower needs of World War II
confirmed the inadequacies of contemporary methods and

1.10

equipment for the training of crews of military air-
craft. Both British and Americans had sought to
eliminate this major weakness by the initiation of
research and development programs designed to create
superior equipment and methods. The programs led to
the operational flight trainer which, without ever
leaving the ground, simulated the flight characteristics
of a particular existing warplane. Such trainers had
proved attractively useful in training flight crews at
inestimable savings in time, money, and lives. The
early British trainer, the "Silloth," was pneumatically
operated. The Americans subsequently applied the same
technique, but when further investigation disclosed that
temperature and humidity variations affected pneumatic
operation too drastically to permit satisfyingly realistic
operation of the complicated systems required in the
trainer, they turned to electrical networks and motor

circuits , and obtained the greater reliability and

3
versatility they desired.

Following the example of the flight trainers , if

a mock airplane cabin or cockpit could be put through

the sort of motions that wind-tunnel tests and calculations

indicated a new and untried design might exhibit, the

responses of a pilot at the mock-up controls would

provide valuable data regarding the promise of the untried

design and, when integrated with further wind-tunnel tests

1.11

and calculations, could effectively accelerate the
development and production of wholly new and superior
airplanes. At least, such was the reasoning of MIT
engineers when they joined in active discussions with
United States Navy personnel in 1943 and 1944.

The Navy planners approached the problem from a
more practical military view. They saw it as an
opportunity to reduce the increasing cost in dollars
and man-hours of proviiding a new and different flight
trainer for each warplane model in combat use. Instead,
a protean, versatile, master ground trainer would be
developed that could be adjusted to simulate the flying
behavior of any one of a number of warplanes . Such a
prototype trainer would provide, they realized, the
configurations and specifications to which cheaper
individual-model trainers might be built in desired
numbers for thexr flyxng schools.

So Navy and MIT engineers, for their separate but
mutually reinforcing reasons , made common cause and in
19 44 embarked on a common project utilizing Navy funds
and MIT technical competence: the development of the
Aircraft Stability and Control Analyzer (ASCA).

As events turned out, ASCA was never built. A
series of consequences that no one foresaw intervened,
Whirlwind I appeared instead, and it was put to a wholly
different use involving the aerial defense of the
continental United States.

1.12

The key figure to set these events in motion was
the Naval Officer who, more than any other one man,
brought the ASCA project into being, Captain Luis de
Florez, director of the Special Devices Division of
the Bureau of Aeronautics. Captain de Florez was one
of those who in the very early days of World War II
had decided to forsake a lucrative civilian career in
order to serve the national cause. An engineer with
an international reputation for his work in aviation
and oil refining, he joined the Navy in 19 39. There,
until his return to civilian life in 1946, he pioneered
in the development of "synthetic" training devices, some
of which one Congressional subcommittee report called
"little short of miraculous."^ In 1944 he received the
Robert J. Collier Trophy of the National Aeronautics
Association for his contributions to the preparation
of combat crews during the Second World War.

In 1940, after flight training at Pensacola, Captain —
then Commander — de Florez was brought to Washington as
special assistant to the head of the Bureau of Aeronautics ,
Vice Admiral John Towers, a friend of long standing. By
April of 1941, the Commander had won recognition for his
advanced training concepts. Later in the same year he
went to London to study British developments in synthetic
training devices, and it was presumably upon that occasion
that he had the opportunity to study the British "Silloth"
trainer.

1.13

Returning to the United States just before Pearl
Harbor, he was placed in charge of a section in the
training division of the Navy. Subsequently, he was
promoted and granted authority to establish the Special
Devices Division with an initial appropriation of $50,000.
Before the year's end the figure had been increased to
$1,500,000; by the end of the following year, 1942, it
had reached $10,000,000. As of November, 1944, the
Division was well established with a staff of some 250
technical officers and 150 enlisted men and civilians.

For Captain de Florez the trainer-analyzer was a
logical and proper extension of existing operational
flight trainers which the Bell Telephone Laboratories
had developed for the Navy, notably trainers for the
PBM, PB4Y2, and F6F aircraft. These simulators permitted
the reproduction of typical operational flight conditions
by means of instrument readings. The instruments within
the cockpit of the trainer were fed data by an "electro-
mechanical computing system" which responded to both
simulated aircraft performance and crew reaction. Suf-
ficient realism was attained to familiarize the flight
crew with the operational characteristics of the type of
aircraft for which they were preparing.

Operational flight trainers were expensive, but
they had proved a technical and practical training success,
It seemed only natural to Navy and Massachusetts Institute

1.14

of Technology planners, prodded by Captain de Flore z,
to extend the concept "into the generalized field of
aircraft simulation" by investigating the feasibility
of a "universal trainer into which constants for
various types of aircraft could be set." 7

During the fall and winter of 1943, Captain de
Florez discussed the dual-purpose simulator with
members of his technical staff and also with repre-
sentatives of the Bell Telephone Laboratories and the
Massachusetts Institute of Technology. Bell's involve-
ment was the obvious consequence of its contemporary
work in operational flight trainers. The Institute's
involvement stemmed from its own personnel's interest
in the problem and from the reputation of its impressive
technical resources. The latter made the Institute a
source of advice and guidance which Captain de Florez
as a graduate found quite natural and easy to tap.
Initially, he had anticipated using the Institute as

a consultant only; the actual engineering development

p
would be performed by the Bell Telephone Laboratories.

While engaged in discussion with Captain de Florez

and his staff, the officers and professors at the

Institute proceeded to expand their own investigations

into the matter. On the eighth of December, N. McL. ("Nat")

Sage, director of the Division of Industrial Cooperation

at MIT, sent off an official letter to Captain de Florez,

1.15

notifying him that the Institute had appointed Professor
John R. Markham as Project Engineer for research on an

Q

"Airplane Stability and Control Analyzer." While the
Special Devices Division investigated the dimensions
of the enterprise that was taking shape, examining
projected costs and identifying industrial laboratories
that might be willing to develop such a trainer-analyzer,
Markham, together with Joseph Bicknell and Otto C. Koppen,
made a study of more detailed technical aspects of the
problem. They drew up a report the following April on
what they called "a proposed method of ensuring satis-
factory handling characteristics of new airplanes," and
circulated it to interested parties. Of particular
significance for the dawning Whirlwind story is their
assertion that "a specialized calculating machine could
be built that could be set up for a particular airplane
according to data obtained by experimental means , and
the pilot's control motions could be fed into the system
by actually having the pilot fly the resulting airplane.

The MIT study was incorporated into the Navy program
as the result of a conference called in January, ISHk to
discuss the feasibility of using the PBM-3 trainer then
under development at the Bell Telephone Laboratories as
the basis for the proposed dual-purpose simulator.
Agreement was reached at the conference to defer further
discussion until specifications had been prepared by the

1.16

MIT group. Once this had been done, discussions would
be resumed, after which the recommended specifications
would be forwarded to the Western Electric Company for
a proposal to be drawn up on the required engineering
work.

By mid-April the MIT report was completed and sent
to the Special Devices Division. It contained the
reasoned conclusion of aeronautical engineering specialists
Markham, Bicknell, and Koppen that it w*s practicable to
design and construct an aircraft control and stability
anaylzer. The success of existing flight trainers,
they noted, permitted the assumption that "a similar
mock-up and calculating machine could be used to develop
the flying characteristics of a projected airplane."
Save for the construction of a flying prototype, the
proposed simulator, they opined, "should provide the
best means of determining flying characteristics of large
airplanes whether the design be conventional or uncon-
ventional." They warned, however, that adoption of the
proposed simulator would require the expansion and

improvement of existing wind tunnel techniques and

12

equipment in order to secure more and superior data.

A copy of the report was sent directly to Captain
de Florez with an accompanying letter from Professor
Jerome C. Hunsaker, head of the Department of Mechanical
Engineering at the Institute. Professor Hunsaker expressed
his conviction that the proposed simulator offered "a

1.17

new tool of very great research significance," permitting
for the first time, if the details could be worked out,
"the controlled motion (handling characteristics) of an
airplane" to be estimated prior to construction. The
heart of the simulator, the analyzer, would be difficult
to design and build, he acknowledged, but the Bell Tele-
phone Laboratories possessed the ability if they would
make the effort. If the Navy undertook the proposal,
the Institute would for its part enthusiastically continue
to cooperate and assist by making available the facilities
of its Wright Brothers Wind Tunnel for the development,
at Institute expense, of the equipment required to

determine "the unusual aerodynamic coefficients needed

13
to feed into the analyzer."

Once the concept had been endorsed by the findings

of the MIT study group, the Special Devices Division

proceeded during the following month to establish a

formal program for its implementation, identifying the

proposed simulator as "Device 2-K, Aircraft Stability and

14
Control Analyzer." On August eleventh, specifications

for both the computer and the cockpit were published,

and the procedures for the selection of a qualified

contractor were instituted. Curiously, the specifications

contained no reference to the use of the simulator as a

master, operational flight trainer, but described it as

a means "to obtain quantitative measurements of the

1.18

stability, control, and handling characteristics of
large multi-engined aircraft" prior to construction,
permitting the distinct inference that if the MIT engineers

had not prepared the specifications, their recommendations

15
had been most influential. The omission, however, was

in no way a reflection of any change in purpose on the

part of Captain de Florez, for during the years that

followed he continued to regard the proposed device as

the prototype of both a master operational flight trainer

and an experimental-aircraft simulator.

Captain de Florez initially had anticipated that

the project would be undertaken jointly by the Bell

Telephone Laboratories and its manufacturing parent,

the Western Electric Company, but ultimately the task

was given to the Massachusetts Institute of Technology.

All in all, some twenty-five commercial and industrial

organizations were considered in the original canvass,

but these either were eliminated or withdrew for various

16
reasons. Apparently, both Bell and Western Electric

were reluctant to undertake the program lest it interfere

17
with Navy contracts of greater immediacy. Furthermore,

by the fall of 19H4, victory was visible over the horizon,

and it is possible that the two companies preferred not

to commit their facilities to a long-term military

responsibility rather remote from their primary peacetime

missions of servicing the needs of their parent organization,

the American Telephone and Telegraph Company.

1.19

Intended or not, the selection of the Institute
was logical and natural. MIT possessed the interest
and the requisite technical resources , and it had
participated in the project from the very beginning.
In addition, Navy negotiators anticipated a substantial
reduction in cost, since the Institute as a non-profit

corporation had lower direct costs and overhead than

19
private industrial organizations. Whatever the

reasons , the Special Devices Division was authorized

in November, 1944 to undertake in conjunction with

MIT a preliminary investigation into the trainer-analyzer.

Captain de Florez's course of action did not go

unchallenged. From the very beginning of his Navy

career his advocacy of technical innovation met criticism,

opposition and even outright hostility, but it is

to be remembered that the history of innovation is also

the history of resistance to change , especially where

institutional officers and custodians are involved.

Since institutions exist to preserve what men value,

they draw some of their strength and substance and

vitality from tradition as well as from innovation. The

opposition to innovation had generally been sincere,

finding its roots, as Elting Morison has noted, in

adherence to the traditional, to the familiar, to the fear

20

of change and of the impact of change upon one's caieer

2 1

if not one's very way of life. When exercised in a

military institution and carried to an extreme, it can,

1.20

as history shows, confound statecraft and endanger the
very security of a modern nation.

Captain de Florez was neither the first nor the
most conspicuous to encounter such resistance while
encouraging technical progress. The First Sea Lord of
the British Admiralty, Sir John Fisher, had encountered
opposition and hostility in the decade preceding the
First World War when he pushed through the dreadnought
construction program and spent millions on the submarine.
Proponents of the German U-boat as an offensive weapon
were unable to win the support of the guiding genius of
German sea power, Admiral von Tirpitz, and thus Germany

neglected to realize the potential of the weapon which

2 2
might have brought her victory in the First World War.

Opprobrium was heaped upon the Board of Ordnance and

Fortification of the United States War Department for

wasting its limited funds on Samuel Langley's unsuccessful

2 3

experiments m heavier- than-air flights. Admiral

William S. Sims, one of the creators of the modern American
Navy, was in constant difficulty because of his support
of innovat ion . ^ 4

The opposition to de Florez' s proposed trainer-analyzer
was sharp and articulate. It was given voice by Captain
W. S. Diehl, Chief of the Aerodynamics and Hydrodynamics
Branch of the Bureau of Aeronautics, who, acting under
oral instructions, had investigated the feasibility and
value of the proposed trainer-analyzer. In his report

1.21

to his superiors, Captain Diehl was bitterly negative,
describing the projected device as "essentially a
physicist's dream and an engineer's nightmare." The
claims made for the simulator were technically unsupportable
and fallacious, Diehl argued. Furthermore, the proposal
was both inappropriate and redundant, since it encroached
upon work already in process under the aegis of the
National Advisory Committee for Aeronautics. These views,
Diehl asserted were shared by other engineers within

both the Navy and the National Advisory Committee for

2 5

Aeronautics .

To counter the adverse criticism voiced by Diehl,
de Florez marshalled his forces within both the Massachu-
setts Institute of Technology and his own organization,
the Special Devices Division of the Bureau of Aeronautics.
The counter-arguments from the Institute study group
reiterated the initial conclusions that the trainer- analyzer
was technically feasible, valid, and of great promise.
Professor Jerome C. Hunsaker, than on leave from the
Institute to serve the National Advisory Committee for
Aeronautics as its chairman , responded in that capacity ,
rejecting the charge that the proposed program would
encroach upon the Committee's work. Instead, Hunsaker
encouraged the Navy to proceed with the project not only
because of its great practical promise, but because the
research was important for itself.

1.22

For his own part, Captain de Florez replied that
the proposed generalized trainer was a natural outgrowth
of the operational flight trainer. It would eliminate
about 80 per cent of the work required for the design
and construction of a specific flight trainer. A
substantial reduction in cost would result at the same
time that a means would be provided to accelerate the
successful design and development of new airplanes.
The criticism voiced by Captain Diehl, he implied, was
just as invalid and unsubstantial as had been earlier
criticism of the projected development of the now
successful operational flight trainer. He recommended.,
therefore, that his Division be authorized to continue

with the project in cooperation with the Massachusetts

2 7

Instxtute of Technology.

Captain de Florez was persuasive. His arguments
were undergirded by a record of demonstrated accomplishment.
On the twenty-eighth of November, Rear Admiral D. D.

Ramsey, chief of the Bureau of Aeronautics, granted the

2 8
requested permission.

Anticipating that approval to continue with the
project and to enter into contractual negotiations with
the Massachusetts Institute of Technology would be forth-
coming, representatives of the Special Devices Division
had met in mid-October with technical and administrative
representatives of the Institute for pre liminary discussions,

1.23

Present at this conference was another who played a key
role in the Whirlwind story, N. McL. ("Nat") Sage. In
his capacity as director of the Division of Industrial
Research, Nat Sage was responsible for the negotiation

and administration of externally-sponsored research and

2 9

development projects conducted by the Institute. From

that office he was to serve as a sympathetic and pro-
tective liaison agent between the project and its Navy
sponsors, as well as between Project Whirlwind and the
MIT administration. Considered by his peers to be an
excellent judge of men, Sage was more apt to support
the man than the project, in the belief fortified by
his experience that a good man meant a good project.
His support of Whirlwind and its leadership was a reflection
of his willingness to aid younger men who had gained his
confidence and respect. It is extremely doubtful
whether Whirlwind could have survived the stormy years
of 1947-1949 had not Nat Sage given it his unswerving
and resourceful support in his dealings both within the
MIT community and with the Navy.

Nat Sage's influence extended beyond the Institute.
Kis was a strong, dynamic personality. His policy views
helped mold the pattern of the relationships that evolved
during the war years between the federal government and
MIT. These relationships were not peculiar to MIT but
were representative of those which developed between
American educational institutions and the government in

1.24

the wartime research and development effort. Sage was
a shrewd and penetrating observer who understood well
the attitudes, the institutional commitments , the
frailties and foibles as well as the strengths and
insights of both the career military minds and the
civilian-in^for-the-duration administrators and contract
officers with whom he had to deal when representing MIT.
Since this wartime cooperation was unprecedented, Sage
had a relatively free hand as he charted unfamiliar seas
in establishing the procedures and forms which were to
guide the contractual relationships between MIT and the
Government. The novelty of these relationships, the
exigencies of the War, and Sage's experience and resource-
fulness cumulatively gave him the power to induce the

Government to accept many of his suggestions concerning

SI

contractual arrangements , and one consequence of this

state of affairs was the broad latitude of options
subsequently made available to the new ASCA project in
the early conduct of its operations. Indeed, by .conser-
vative institutional and corporate standards the project
enjoyed greater freedom of operational choice than many
responsible executives find it comfortable to contemplate
allowing their enthusiastic younger subordinates.

Another influential MIT representative present at

the October,, .'1944 discussions was Professor Gordon S.

• • 3 2

Brown, the .director of the Servomechanisms Laboratory.

1.25

Professor Brown's presence at the discussions with the
Special Devices Division indicated that if the Institute
chose to proceed with the next phase of the ASCA project,
the Servomechanisms Laboratory might well be involved.
This was understandable , for the nature of the work lay
within the competence and experience of the Laboratory.
The Servomechanisms Laboratory had been established in
December, 1940, under the direction of Gordon S. Brown,
assisted by Albert C. Hall, John 0. Silvey, and Jay W.
Forrester. It was the outgrowth both of a training
program for United States Naval Fire Control Officers
begun in 19 39 in the Department of Electrical Engineering
and of arrangements made by the Sperry Gyroscope Company
with the Institute to undertake a research and development
program that would produce a remote control system for
antiaircraft guns on merchant ships .

Effective defenses were needed against Nazi dive
bombers, which in the fall and winter of 1940-1941 had
become a primary menace to the supply ships approaching
the United Kingdom from the United States and elsewhere.
Necessary to a particular defense system under develop-
ment by the Sperry Gyroscope Company was a servomechanism
that would link a computing sight to the 37 mm. guns with
which merchant vessels were to be armed. Rather than
retool to manufacture an already existing British remote-
control system, the company had chosen to develop a
system which would utilize to the greatest possible extent

1.26
components already in domestic production. To this end

the Company had arranged with the Massachusetts Institute

3 3
of Technology to conduct the necessary research.

Within the Institute, the responsibility for the re-
search program was given to the Department of Electrical
Engineering because of its experience in servomecha-
nisms; in turn the Department organized the Servomech-
anisms Laboratory.

From the beginning the new laboratory was a
loosely controlled organization, for it played a
very special role in Professor Brown's thinking.
Believing that the conduct of research and develop-
ment under very liberal controls was essential , he
refused to employ the procedural controls that many
would have considered mandatory features of good
mangement practice. From a conservative critic's
point of view, Brown provided a dangerously decen-
tralized "every man for himself" environment allowing
too great autonomy to be practical and safely business-
like. It permitted each project director within the
Laboratory to organize the work according to his in-
dividual peculiarities and capabilities. Carried to
the next logical step, it left to each investigator all
the latitude he could wish for in the conduct of his
work. If the man's talents were not up to the task to
be performed, this latitude permitted deficiencies to
become quickly apparent.

1.27

From Brown's point of view, it was a matter of
finding a good man and backing him by turning him loose
to make his own mistakes. In the case of the Servo-
mechanisms Laboratory on the MIT campus , the good man
preferably took the form of any brilliant and promising
graduate student in electrical engineering who gave
indications of being able to avoid the gross mistakes
and of profiting rapidly from the small ones. Brown's
surveillance was perhaps deceptively loose because he
gave his project directors such wide leeway. The more
astute students soon realized that this procedure gave
them all the rope they needed to hang themselves as
high and spectacularly as one could wish, and one
effect of this realization was the exercise of prudent
caution and more careful planning while being innovative.

The unconventional management techniques and pro-
cedures Brown applied were so inconspicuous as to seem
almost absent. Some of his own subordinates in the
Laboratory became convinced that he really did not
know what was going on, so often was his back apparently
turned. This apparently casual supervision was de-
liberate, however, reflecting Brown's philosophy of
education and his ideas on the proper conduct of ad-
vanced research and development. Brown was convinced
that the loosely structured but, for his purposes,
highly communicative interchange of ideas and problems

1.28

which resulted not only contributed to the growing
maturity of the student but also enabled the older
faculty members involved to remain more innovative
and more critical of their own technical views. A
net result would be the more rapid and sound progress
of engineering knowledge , for regardless of academic
level, professors and students alike were stimulated
by their mutual contacts and exchanges of views in
this informal research-laboratory environment.

Brown felt then and in after years that successful
and original engineering research could more likely
be achieved if the research and development problem
were pursued by students caught up in an instructional
program. It was not enough to provide the intellectual
milieu, the intellectual challenges, the new horizons
that a first-rate educational and training program
could offer. Necessary preliminaries as these were,
they were too protectively academic. The harsher,
more realistic practical experience of the bona fide
research and development laboratory committed to
solving non-academic problems was also necessary, nor
should such experiences be postponed until after
graduate degrees had been obtained. Brown saw no reason
why carefully selected predoctoral and premaster's
degree students of the caliber that HIT attracted
should not be exposed to the novel blend of the sheltered,

1.29

academic instructional program and the playing- for-
keeps , practical, research and development program that
they would encounter during the remainder of their
experience as professional engineers. In his direction

of the Servomechanisms Laboratory, Professor Brown sought

3 5
to implement these convictions. The measure of his

success was demonstrated not only by the considerable
performance of the Laboratory itself, but also by the
performances of former students and assistants in
later years .

The spirit of the Laboratory was high, in part
because it was the product of Professor Brown's
inconspicuous leadership, but also in part because
other factors operated. One was the glan of the
graduate student and research assistant who, having
embarked upon his professional career, is determined
to demonstrate his creative abilities and competence
and to find new worlds to conquer. This glan Professor
Brown sought to further and exploit. Another factor,
equally strong, was the personal dedication the War
evoked. Whether this sense of personal commitment
stemmed from pure patriotism or the desire to get a
"dirty" job done, it was as much a stimulant to the young
neophyte in the Laboratory as it was to his senior
mentors and colleagues of the scientific and engi-
neering community. After the Japanese attacked Pearl

1.30

Harbor on December 7, 19H1, the American people com-
mitted themselves wholly to the war effort, and there
arose a national mood of determination and self-
sacrifice difficult to imagine and reconstruct in
all its intensity by those who have not experienced
it. It became a force whose impact upon every citizen
was not lightly to be discounted, and the response to
the nation's call when it mobilized its scientific
and engineering manpower to aid the prosecution of
the war attests to the power of this mood. The ur-
gencies of the War - to many, the conflict was the very
battle for survival of the American way of life- made
it difficult if not impossible to adhere to a "business
as usual" philosophy.

The operational latitude within the Servo-
mechanisms Laboratory encouraged the exercise of
these motivations and enthusiasms. Much of the same
psychological atmosphere, the same 61an, the same
personal response were to be carried over into
Project Whirlwind, — and years later were recalled
with longing and nostalgia by those who had been
participants .

In the years following its establishment, the
Servomechanisms Laboratory had expanded both in programs
and in personnel. By the time MIT was discussing the
ASCA project with the Special Devices Division, the
Laboratory had a staff of approximately 100, including

1.31

thirty- five engineers. It had since its creation

"developed remote control systems for 40mm gun drives;

for radar ship antenna drives ; for airborne radar and

turret equipment ; and for stabilized antennas , directors

and gun mounts; as well as having cooperated in a

number of other instrument problems." 36 As a consequence,

the Laboratory had in its four years acquired extensive

experience in the research, design, development, and

practical test of that general class of machines,

an example of which it was anticipated would form

the heart and brain of the projected trainer-analyzer.

During the month which followed the October
conference between the Institute and the Navy, both the
Special Devices Division and the Servomechanisms
Laboratory sought to arrive at an unofficial under-
standing which could serve as the basis for official
contractual negotiations between the Navy and MIT. A
tentative proposal was prepared by the Laboratory in
early November, providing for a research and develop-
ment program which would be carried to the "breadboard
model" stage over a one-year period at an estimated
cost of $200,000. The construction of the final
simulator would be undertaken only after the program
had then been re-evaluated and the decision to
continue had been made. 37 A conference held on
November 15th disclosed, however, that both parties

1.32

had come to the opinion that the initial proposal was
both too extensive and too expensive. Consequently,
they jointly worked out a new proposal, recommending a
more modest preliminary study which would cost about
$75,000. This study would provide, they felt, a more
accurate appraisal of the feasibility and ultimate cost
of the trainer-analyzer.

The terms of this agreement were incorporated in the
Special Devices Division's application to the Bureau of
Aeronautics for approval of the project. On December It,
19f4, the Navy issued a formal Letter of Intent for
Contract Noa(s)-5216. Four days later the Institute
officially accepted the Letter, and the program for the

development of the Airplane Stability and Control

3 8
Analyzer was officially launched. None of the partici-
pants anticipated a major change in course. In this
they were quite reasonable and quite wrong, for no
one could anticipate, before the research was undertaken,
that the difficulties inherent in realizing the initial
purpose would be so profound or that the efforts of
both MIT and Navy experts to reach a solution would
generate a different enterprise superseding the first.
Even the prophetic Captain Diehl, who had called the
project "a physicist's dream and an engineer's night-
mare," did not allow for a change in course; after all,
his solution had been to refrain from embarking on it
at all.

NOTES TO CHAPTER 1.

1. Vannevur Bush, Modern Arms and Free Men (New York,
1949) , pp. 6-7.

2. James McCormack and Vincent A. Fulmer, "Federal
Sponsorship of University Research," The Federal
Government and Higher Education , The American As-
sembly, Columbia University (Englewood Cliffs, N.J.,
1960), p. 78.

3. Draft memo, (anonymous, no date). The contents sug-
gest a Navy source prepared it prior to April, 1944.

4. SDD Memo, Oct. 11, 1944, J. W. Ludwig to Capt . de
Florez. Cf. Enclosure "D" of BuAer Procurement
Directive EN 11-27339-45, Nov. 22, 1944: "Data for
OPSM," Nov. 20, 1944, by J. B. Van Duzer and R. I.
Knapp of SDD.

5. Quoted in Robert L. Taylor, "Captain Among the Syn-
thetics," The New Yorker , Nov. 11, 1944, p. 34.

6. Robert L. Taylor, "Captain Among the Synthetics,"
The New Yorker , Part I, Nov. 11, 1944, pp. 34ff ;
Part II, Nov. 18, 1944, pp. 32ff; see also Luis de
Florez' s obituary in The New York Times , Dec. 6,
1962, p. 43.

7. Servomechanisms Laboratory, MIT, Project Whirlwind ,
Summary Report No . 1 (Apr. , 1946), pp. 1-4.

8. Memorandum, Rogers Follansbee , Aircraft Simulation
Section, to Director, SDD, subj . : "Analyzer, Flight
Characteristics," Feb. 5, 1944.

9. Ltr. , N. McL. Sage to Capt. Luis de Florez, Dec. 8,
1943.

10. John R. Markham, Otto C. Koppen, Joseph Bicknell,
Note on A Proposed Method of Ensuring Satisfactory
Handling Characteristics of New Airplanes , (April ,
1944), p. 4.

11. Ltr., Rogers Follansbee, Aircraft Simulation Section,
SDD, to J. Bicknell, MIT, subj.: "PBM-3 Operational
Flight Trainer Data-Forwarding of," Feb. 5, 1944i

12. Markham, Koppen, Bicknell, Note on A Proposed
Method . . . , pp. 1-6.

13. Ltr. , J. C. Hunsaker to Capt. Luis de Florez, Apr.
15, 1944.

14. (C. P. Andrade), Memorandum for Files, subj.:
"Project Whirlwind," June 13, 1946.

15. Navy Dept. , BuAer, SDD, Specifications for Airplane
Stability and Control Analyzer , Aug. 11, 1944.

16. Memorandum, Head of Production Branch, SDD, to Direc-
tor, SDD, subj.: "Project 2-K - Report on Companies
Considered and Facilities Available," Oct. 13, 1944;
Ltr., L. F. Jones Gov't Development Section, RCA, to
R. I. Knapp, SDD, Oct. 18, 1944; Ltr., W. S. Hill,
Ass't District Engineer, General Electric Co., to J.
B. Van Duzer, SDD, Oct. 11, 1944.

17. Memorandum J. B. Van Duzer, SDD, to E. N. Howell, SDD,
subj.: "Sources for Project 2-K, Stability Control
Analyzer and F7F OFT's," Sept. 18, 1944.

18. Enclosure "D" of BuAer Procurement Directive EN11-27339-
45, Nov. 22, 1944; "Data for OPSM," Nov. 20, 1944

by J. B. Van Duzer and R. I. Knapp of SDD.

19. Memorandum, Luis de Florez to Rear Admiral D. C.
Ramsey, BuAer, Nov. 27, 1944.

20. Memoj, Director, SDD, to Chief, BuAer, subj.: "Air-
plane Stability and Control Analyzer," Oct. 13, 1944.

.21. See Professor Morison's essay, "Gunfire at Sea: A
Case Study of Innovation," published in his Men ,
Machines , and Modern Times (Cambridge , Mass . , 1966 ) ,
pp. 17-44.

22. A. J. Marder, From the Dreadnought to Scapa Flow,
vol. 1, The Road to War, 1904-1914 (London, 1961),
pp. 330-335.

23. Mark Sullivan, Our Times , The United States 1900 -
1925 (New York, 1932), II, America Finding Herself ,
pp. 557-568.

24. Elting Morison, Admiral Sims and the Modern American
Navy (Boston, 1942 ) , passim .

25. Memorandum, Head of Aerodynamics and Hydrodynamics
Branch to Chief, BuAer, sub j . : "Airplane Stability
and Control Analyzer — Comment on," Sept. 5, 1944 (En-
closure "A" to ltr. , Director, SDD to Chief, BuAer,
subj . : "Airplane Stability and Control Analyzer,"
Oct. 13, 1944.

26. Memo, Comments on Captain Diehl's Letter with Regard
to ASCA (MIT memo), Oct. 2, 1944; ltr., J. C. Hunsaker
to Capt. Luis de Florez, Oct. 4, 1944 (Enclosures

"C" and "b; t respectively, to ltr., Director, SDD, to
Chief, BuAer, subj.: "Airplane Stability and Control
Analyzer," Oct. 13, 1944).

27. Ltr., Director, SDD, to Chief, BuAer, subj.: "Air-
plane Stability and Control Analyzer," Oct. 13, 1944.
Cf. Enclosure "E" Memorandum, J. W. Ludwig to Capt.
de Florez, subj.: "Analysis of Project 2-K Airplane
Stability and Control Analyzer," Oct. 11, 1944, ibid .

28. Approval initialled on Memorandum, Capt. Luis de
Florez to Rear Admiral D. C. Ramsey, BuAer, Nov. 27,
1944.

29. Memorandum, J. B. Van Duzer, subj.: "Project 2-K,
Aircraft Stability Control Analyzer, Conference with
MIT representatives," Oct. 18, 1944.

30. Interview, Prof. G. S. Brown, MIT, by the authors,
Jul. 6, 1964; Interview, Prof. J. W. Forrester, MIT,
by the authors, Jul. 24, 19 64.

31. Ibid .

32. Memorandum, J. B. Van Duzer, subj.: "Project 2-K,
Aircraft Stability Control Analyzer, Conference with
MIT representatives," Oct. 18, 1944.

33. Interview, Jay W. Forrester and Robert R. Everett by
the authors, Jul. 31, 196 3.

34. Jay W. Forrester, "Hydraulic Servomechanism Develop-
ments," MS Thesis, Dep't of Electrical Engineering,
MIT, June, 1945, pp. 1-3.

35. Interviews by the authors with: J. W. Forrester and
Robert R. Everett, Jul. 31, 196 3; Kenneth H. Olsen,
June 24, 1964; Charles W. Adams and John F. Gilmore,
Jul. 3, 1964; Gordon S. Brown, Jul. 6, 1964.

36. N. McL. Sage, Director, DIC, MIT, to Chief, BuAer,
att'n.: Lt. Comdr. E. N.' Howell, SDD, subj.:
"Proposal for Contract for Development of a General-
ized Multi-Engined Operational Flight Trainer,"

May 22, 1945 •, draft of ltr. from G. S. Brown to
Chief, BuAer, att'n.: Lt . J. B. Van Deusen, SDD,
subj . : "Proposal for Contract for Development of
Aircraft Analyzer," Nov. 3, 1944.

37. J. W. Forrester, MIT Computation Book, #36, p. 14;
draft ltr., G. S. Brown to Chief BuAer, att'n, J.
B. Van Deusen, subj.: "Proposal for Contract for
Development of Aircraft Analyzer," Nov. 3, 1944.

38. Enclosure "D" of BuAer Procurement Directive EN11-
27339-45, Nov. 22, 1944: "Date for 0P8M," Nov. 20,
1944, by J. B. Van Deusen and R. I. Knapp, SDD; J.
W. Forrester, Administrative notes entered in his
MIT Computation Book #36, p. 14; Memorandum, Luis
de Florez to Rear Admiral D. C. Ramsey, BuAer, Nov.
27, 1944; Navy Dep't., BuAer, Letter of Intent for
Contract NOA(s)-5216, Dec. 14, 1944.

COMPUTING PROBLEMS EMERGE

A natural change in the course of the investigation
occurred as a consequence of the preliminary aerodynamic
analyses of the Airplane Stability and Control Analyzer's
prospects and problems. It occurred between the time
Captain Luis de Florez had initiated preliminary dis-
cussions with the Massachusetts Institute of Technology
in 19 4-3 and the time the Institute accepted the Letter
of Intent over a year later. It was set in motion
when de Florez asked the Institute how practical his
ASCA project to build a simulator appeared to be from an
aerodynamics viewpoint. As we have seen, the response
of Professor Hunsaker and the Wright Brothers Wind
Tunnel engineers was that the problem appeared by no
means insoluble, and the subsequent, more detailed
investigation and conclusions of Markham and his
associates reaffirmed the reasonableness of de Florez'
proposal and indicated it was attractively worth
further consideration.

The consequence of these conclusions was a shift
in the focus of investigation from the field of
aerodynamics and aeronautical engineering to the field
of electrical engineering and electromechanical control
systems. So the problem passed to Professor Gordon
Brown and the Servomechanisms Laboratory. The formal
agreement reached in December, 1944 occurred,

2.1

2.02

of course, after the fact of Brown's involvement, and
it betokened not his decision to become involved but
his commitment and that of his engineers to pursue the
ASCA project and its electromechanical simulation problem
further. He had already brought the problem to the
attention of one of his assistant directors, Jay W.
Forrester, who had managed earlier projects in the
Laboratory. Forrester became interested in this
provocative engineering challenge, as Brown had hoped,
and accepted the direction of the ASCA project in the
fall of 1944. It was a responsibility that he was
not to relinquish until 19 56.

While remaining in charge, Forrester soon brought
Robert R. Everett into the project. In a very special
way, reflecting the complementary temperaments of the
two young men, Everett came to share the responsibility
and the technical direction of the project with Forrester.
These were the two engineers whose technical and
administrative leadership gave the project its basic
character during the following decade. There was
never any question that Forrester was in charge of the
project, exercising administrative authority and
technical leadership, and there was never any question
that Everett was second in command, exercising con-
tinuing technical leadership and administrative authority
when Forrester was preoccupied with external affairs.

2.03

Linked by a deep mutual respect and understanding,
they worked together in unusual harmony, without
always employing the same means to reach their common
goal.

A native of Anselmo, Nebraska, Forrester had
obtained his Bachelor of Science degree in engineering
at the University of Nebraska in 19 39. In the fall of
that eventful year (World War II had begun when
Hitler invaded Poland in September, 19 39) Forrester
came to MIT as a graduate student and research
assistant in electrical engineering. He was already
on hand when Brown set up the Servomechanisms
Laboratory in response to the looming technical
demands of the war. The progress of the war expanded,
the opportunities for original engineering research
at MIT by providing the incentive , the needs , and
the funds. One of the research and development fields
so expanded and accelerated involved the design and dev-
elopment of feedback circuits and mechanical and
electrical analogue devices and powerful servomechanisms
responsive to remote control. It was a field that saw
dramatic technical progress during the war, and since
the "Servomech Lab" was in the middle of it, Forrester
was one of those who acquired extensive familiarity
with the potentialities and the limitations of servo-
mechanisms and with associated problems of integrated

2.04

system design and development. Fighter-director
radar controls later placed on the USS Lexington
were one of the systems that had given him important
practical experience. Consequently, when de Florez'
trainer-analyzer appeared above Forrester's horizon, he
possessed both the technical experience and the
administrative organizational experience to set up
the project. Because of the unpredictable character of
the research and development process , neither he nor
anyone else at the time realized what the project would
become and what transformations would ensue during the
following eighteen months, not to mention ten years.

Everett, born in Yonkers , New York, had received
his B. S. degree in electrical engineering at Duke
University in June, 19H2, six months after the United
States became a combatant in the war. In the summer
of that year, about a month after entering MIT to
seek a master's degree, Everett joined the war effort
by going to work for Forrester in the Servomechanisms
Laboratory.

Both young men thus were exposed to Brown's way
of doing things and to the level of intellectual enter-
prise maintained by him and his colleagues. Under his
eye they developed their respective organizational and
administrative talents as well as their electrical
engineering expertise. Although they did not try to

2.05

duplicate Professor Brown's personal style, it is not
surprising that features of the philosophy of management
followed by Brown within the Servomechanism Laboratory
influenced significantly the organization and admin-
istration of the Airplane Stability and Control Analyzer
program after it became Forrester's primary responsi-
bility. He and Everett proceeded, of course, to conduct

9

the program xn thexr own style.

After accepting technical and administrative
responsibility from Brown, Forrester worked on the
ASCA project virtually alone at first and by the end of
the first week in November had laid out his plan of attack.
Basic units of the complete analyzer would include a
simulator "cockpit with controls and instruments, the
flight engineer-observer station, and the calculating
equipment." While the specifications seemed to have
purely electrical analogue computing in mind, Forrester
surmised that many of the integrator functions "might
well be met through use of a variable-stroke hydraulic
transmission." Perhaps a mixed mechanical and synchro
data system* although more expensive, might avoid
certain design difficulties of the all-electric system.
"A combination system of synchro data, voltage data,
mechanical integrators for multiplication by constants,
and hydraulic transmission integrators for integrating

and for multiplication by two variables" might be a

3

suxtable compromxse.

2.06

Obviously, he should study existing trainers,
familiarize himself further with the equations
embodying the aerodynamic requirements , "discuss the
objective of the apparatus with the Navy sponsors. . .
with commercial test pilots, designers, and wind
tunnel men for detailed information on behavior and
accuracy," examine "mechanical and electrical methods
of continuous mathematical calculating," obtain engine-
performance equations, study the physical details
involving "types of signalling [and] types of amplifiers
and other components," consider the types of schematic
approaches available, and lay out a schematic solution
that would "reduce the number and types of equipment
as far as possible."

On November 4th and 5th he laid out preliminary
schematics "to show the solution of the equations"
contained in the specifications. Familiarizing
himself further in this way with the abstract statements,
conditions , and quantities that the Airplane Analyzer
would translate into suitable motions of the simulator
cockpit, he considered ways and means of interpreting
and restating for engineering purposes the requirements
set forth in April at MIT by Markham, Koppen, and
Bicknell and in August by the Bureau of Aeronautics.
His preliminary survey indicated that there were
ninety-two quantities and thirty-three simultaneous

2.07

equations involved, just to describe the aircraft
response. Further study indicated that, strictly
speaking, thirty of the equations described the air-
craft response, three related to acceleration and
velocity, eight dealt with instrument responses, and
six applied to the control forces. So the Analyzer
would have to handle at least 47 equations involving
5 3 variables with respect to time, and none of these
took into account the engines and engine controls.
Since a multi-engine simulator was what de Florez
had in mind, the device would be complex, indeed.

From his preliminary schematics Forrester further
drew the regretful conclusion that "the extensive
use of synchro position for quantities or of mechanical
multiplication seems entirely out of the question."
On the other hand, "a-c voltage signals should cause
much less difficulty because of the ease of isolating
various circuits." Careful engineering, he noted,
ought to be able to avoid phase difficulties such as
Bell Telephone Laboratories engineers had encountered
when designing trainers.

In this manner he proceeded to shape his pre-
liminary assessment of the ASCA problem. It was
partly on the basis of this assessment that the
meeting of November 15, 1944, between Navy Bureau of
Aeronautics personnel and Sage, Brown, and ASCA

2.08

engineers Forrester, Everett, and Hugh Boyd of the
Institute called to discuss contract arrangements found both

the Navy and MIT representatives ready to back away
from the $200,000 bread-board-model contract that had
been proposed earlier. As Forrester noted at the time,
"BuAer felt from previous projects that the project
would not be of such magnitude and also , after the
intervening two weeks of study, MIT had a clearer
picture of the requirements . " It appeared more practical
to think in terms of "a 4 to 6 months preliminary
study to be covered by an appropriation of $75 , 000. ... " 7

What had happened was this: MIT, through the
informal actions of Sage, Brown, and Forrester, had
initiated the limited feasibility and cost study
stipulated in Contract N0a(s)-5216 even before the
Letter of Intent was issued by the Navy and accepted
by the Institute. This was neither the first nor the
last time that professional involvement with the
engineering problems by the engineers preceded official
endorsement by the appropriate administrative and
legal officers. Indeed, it was the practical thing
to do: assess the problem in a preliminary way
before making a commitment to undertake it in greater
detail. In a very real sense Forrester's work before
December was a feasibility study of the prospect of
taking on the ASCA feasibility study.

2.09

This arrangement a not infrequent characteristic

of the research and development process allowed

formal fiscal and administrative agreements to rest
upon the latest technical thinking and had the merit
of placing the entire procedure on a more empirically
sound basis attractive to all the parties concerned.
In consequence, the legal and monetary relationships
became subsidiary means to the end of securing the
engineering knowledge and the technical hardware
sought. As will be seen, this subordination of fiscal
and administrative factors to the engineering factors
did not persist throughout the history of the project.
But it was a customary way to start a project and
quite acceptable in view of the fact that the war was
still on. While prudent control of expenditures was
always to be desired, cost itself was no object; the
imperative consideration was to get on with the job,
at whatever the cost in dollars. In such a wartime
policy climate (the only kind in which Forrester had
accumulated his research and development experience),
it was natural that fiscal and administrative policies
should be subordinated to the technical needs of those
who were getting the job done.

Forrester's investigative techniques were, of
course, the product of his experience acquired since
at least 1940. They were not intuitive, unexamined

2.10

procedures that he was unaware of and could not explain.
On the contrary, his was a temperament that took it for
granted he should analyze and make as explicit as possible
the useful techniques that "came naturally" from his
experiences. His was a mind that preferred to know
where it stood and why, at all times. It was committed
in a very self-aware way to understanding and ration-
alizing and systematizing the intellectual procedures
through which it moved, especially where innovative
activity, such as engineering research evoked, was
involved. This trait was part of the young graduate
student's immense self-possession (that some found
presumptuous, if not patronizing, in one so youthful).
It helped him to organize his plans of attack, it
helped him to carry them out, and although it did not
prevent errors in judgment, it provided continuing
re-examination of that judgment and helped to minimize
errors before they got out of hand. It could not
forestall basic policy- level errors, nor was it a
remedy for the fact that the fullness of his expert
knowledge in the area of mechanized analog computation
and the principles of servomechanisms was also the
measure of the depth of his contemporary ignorance
of mechanized digital computation, resulting in a
postponement in his selection of a suitable computer
while he endured progressive disenchantment with the
ideal device in his mind's eye that he had at first
selected.

2.11

An example of his self-aware, analytical mode of
procedure is to be seen in. a report, his master's thesis,
which he began in 1941 and finished in November, 1944,
as he was taking up the ASCA project. The views he
had expressed in 1941 he considered still appropriate
in 1944. In discussing the scope of the thesis, he
made no apology for the fact that "considerable
emphasis is given to the mathematical analysis of the
control systems which have been developed." "This
has not been done because the analysis is academically
fascinating," he wrote, "but because, from an engineering
viewpoint, it has proven the surest and quickest
way to obtain the desired results and to avoid the pitfalls
so often appearing in the trial and error attempt to
solve a complex problem." Forrester felt that the
analysis of the specific servomechanism discussed in
his thesis provided "an excellent example of the
philosophy of the laboratory toward remote control

o

theory. "

"It may seem," he continued, "that an undue amount
of attention is devoted to the development and design
of the early experimental and pilot models. However,
it is there that the analytical approach may most
effectively be shown, and the brief dismissal of many
of the design and engineering problems of later work
results not because these problems were easily solved,

2.12

but because one with the necessary understanding and
respect for the complexity of the operating principles

q

may expect to reach the proper answer."

Forrester regarded it as a telling virtue that
a "great deal of time and attention is devoted at the
Servomechanisms Laboratory to careful measurements of
the characteristics of individual pieces of equipment
which are to be placed in a remote control system.
These measurements and study yield information on the
reliability of the components and make available
numerical values of the constants appearing in the equations
representing the response of a system. Such an in-
vestment of time and effort has returned substantial
and satisfying dividends in the reduction of time
consumed by ' cut-and-try ' experimenting."

Such were the technical background and perspectives
that Forrester brought to bear when he opered his invest-
igation of the ASCA problem in November, 1914. By
mid-December he had become sufficiently acquainted
with both operational flight trainers in general and
the proposed trainer-analyzer in particular to arti-
culate the technical requirements for the analyzer,
prepare a tentative time schedule, and assemble a
list of personnel whom he considered competent to
carry out the work needed in fulfillment of the contract's
aims and terms .

2.13

Among the personnel he sought, in addition to
Everett , were three other engineers : Hugh Boyd ,
Stephen Dodd, and George Schwartz, all of whom had
been working on projects in the Servomechanisms
Laboratory. As was to be expected, however, some
time elapsed before Forrester was able to assemble
the engineering staff he wanted. By the following
February only Forrester himself and Boyd had been
able to devote full time to the project; the others
divided their time between the trainer-analyzer and
other projects within the Laboratory .

Work on the project did get under way, nevertheless,
and at the start it was paper work. Early in March,
a five-page laboratory report outlining methods of

mounting and actuating the simulator cockpit was given

12
to Forrester by one of his engineers. The cockpit

problem appeared to admit more of straightforward
solutions than did the prospect of designing the ex-
tremely complicated analog computer that the simulator
would require. As more problems and subproblems were
investigated, even more problems were uncovered, with the
result that as the six-month preliminary study period
progressed, Forrester became less sanguine than he had
been. The problems were proving more formidable than
he and his associates had thought in October and
November that they would be. Overall, however, the

2.14

situation was under control, for the point of such
research was to delineate the scope of the problems
involved, and the emerging picture continued to in-
dicate that the stumbling blocks were not insurmountable,
Forrester, Brown, and Sage remained confident they
could achieve a solution, although by May, 1915, they
were willing to admit that they had, in their earlier,

relative ignorance, justifiably underestimated the

13
development cost.

In a memorandum of May 8th that never went

beyond the draft stage, Forrester noted that the problem

of developing certain components for the Analyzer was

requiring more extensive research than they had

14
expected. Many of the customary electrical and

mechanical procedures of solving the appropriate

differential equations could not be applied in their

usual form but would have to be improved and tailored

to the job at hand. Particularly was this true,

Forrester felt, where speed of response was critical

and where the ratio of maximum to minimum signal was

extremely broad. In the forme -1 " instance, the reactions

of the pilot in the simulator cockpit would enter into

the statement and the solution of the hypothetical

aircraft's stability. His responses to simulated

aerodynamic forces acting upon the pilot's controls

ought to accomplish corrective actions , and these

would have to take effect as promptly in the simulator

2.15

as they would in a flying airplane. The scale of
allowable response times was limited to the time it
would take in an actual airplane. The equipment had
to operate within these real response times. "This
is especially true," wrote Forrester, "of the inte-
grators which convert accelerations to velocities
and of the control-column loading equipment."

Again, both the normal maneuverability of an air-
plane and its range from smooth, level flight to sharp
maneuvers imposed a corresponding range between min-
imum and maximum signals that no known mechanical
equipment having a single scale of operation could
embrace unless, by suitable mechanical and electrical
means, one incorporated "an automatically self-adjusting
scale factor. Details of variable scale-factor devices
have been worked out but have not yet been experimentally
proven. "

Consequently, when Forrester contemplated the end
of Phase 1 (the preliminary study period) and the
beginning of Phase 2 (actual design and construction of
ASCA itself), as these were called for in the contract,
he recognized that they could not know where they stood
in Phase 1 until the studies and demonstrations then
in progress were completed, nor could they provide a
realistic answer regarding the practicality of Phase 2
until they knew that suitable components existed. He

2.16

felt that the components could be developed "in the
next few weeks," but saw the necessity of obtaining
additional laboratory and engineering time. Perhaps
an extension of time on Phase 1 and scrutiny of the
transition period between Phase 1 and Phase 2 would
constitute the best course of action. If $50,000
were added to the original $75,000 allotted for Phase 1,
a sufficient extension of time might be achieved.
An interval of two or three months, at $25,0 00 per

-1 c

month, might take care of the transition period .

This precise course of action was not taken. The
Navy remained satisfied with the rate of progress
made during the first five months. It was confident
that the "general outlook was promising." When the
Institute's Division of Industrial Cooperation--
Sage's office — submitted on May 22, 19 45 its proposal
for extension and modification of the contract, the
upward revision of estimated cost that Forrester, Brown,
and Sage felt was necessary omitted the complicated
phasing of phases that Forrester had toyed with and
stated instead that the project could be carried through
to completion within eighteen months at a cost of
approximately $875, 000. 17

The Navy's response was to renew and continue the
project under Letter of Intent for Contract NOa(s)-
70 82, dated June 30, 1945. By this renewal the
Navy firmly committed itself to the project, for the

2.17

new contract not only continued, but expanded the
project at a cost which was too large to permit easy
withdrawal. Since the aims of de Florez and his
assistants and the aims of the MIT personnel were in
fundamental agreement — to build an Aircraft Stability
and Control Analyzer — , since the Institute had become
well committed, first through the efforts of Sage and
Hunsaker, then through the efforts of Sage and Brown,
and now through the efforts of Sage, Brown, and
Brown's competent assistant, Forrester, and since ex-
cellent contractual and working relationships on the
technical level had been established, there was every

reason to go ahead. Of course, there were unsolved

18
problems. Had there been no problems, the Navy would

not have had to turn to the Institute, and one of the
private manufacturers less interested in advanced
research could have taken on the job. As to the
immediate future, no one could say how long it would
take to finish the war against Japan; the empire the
Japanese had begun to build in 19 32 might be crumbling
rapidly, but no one could be sure how long the suc-
cessful invasion of Japan itself would take. And if
ASCA were developed too late to help in the War, its
long-run achievement would still be useful. Eighteen
months and $875,000 would represent a sound prospective
investment of Navy research and develop ment funds for
1945 and 1946.

2.1J

The onset of summer saw increasingly severe
technical analysis by Forrester and his associates of
the progress they were making and the problems they
were encountering. Forrester in particular began
to probe insistently the problems of the form the
computer equipment should take and the appropriate
representation that should be given the test data. To
represent nonlinear data by mechanical linkage, for
example, appeared in prospect to be neither a flexible
nor a general enough method. Although it could be
incorporated in an analog system, it posed the possi-
bility that wasteful trial-and-error routines would

have to be undertaken each time new data required

19
adjustment of the lmakge system.

Forrester discussed his problems with others.

Professor Samuel H. Caldwell of the Electrical

Engineering Department had suggested in Hay that

the work of George R. Stibitz and his associates at

Bell Telephone Laboratories might offer suitable

alternatives , but Forrester did not pursue this lead

20 . .
at the time. Stibitz, a mathematician who had ob-
tained his Ph.D. in physics, was then involved in the
design of a digital computer using telephone relays
for storage of numerical data and for arithmetical

operations. The following year the Bell Relay Corn-
on
puter, Model 5, was put into operation.

2.19

By the last week of June, 19 45 Forrester was
notifying Brown that the rough survey of requirements
his group had made in anticipation of the Navy's con-
tinued support indicated they needed greater manpower.
Eight more electrical and electronic engineers and
three mechanical engineers (they had none at present)
were needed. In addition, a building would have to
be designed and built; consequently, an architect
should be obtained to supervise construction through
the following spring. Forrester went on to outline a
schedule of the progress required to carry out Phase
2: research from August, 1945 until the first of the
following year, mechanical development and design from
August, 1945 to March, 1946, electrical development
and design from January to March, 1946, procurement
and construction from January to July, 1946, assembly
and installation from July to December, 1946, testing

and trouble-shooting from January to March, 1947, and

22
delivery of the equipment to the Navy on March 31, 1947.

As June passed into July and July into August
during that summer of 1945, Forrester became increasingly
disenchanted with the lack of flexibility and versa-
tility of the elaborate servomechanism system that
was taking tentative shape. The real-time response
problem still defied forthright solution, and unless
certain design features were changed, the units of the

2.2Q

Analyzer would remain permanently interconnected in the
pattern imposed by the equations of motion of the air-
craft. In consequence, the computer portion of the
Analyzer would be unavailable for use on other problems
between simulation tests. The result would be a grossly
uneconomical waste of potentially one of the most power-
ful such computers to come into existence. Perhaps
pre-wired, removable plugboards could be employed, with
the result that operating characteristics of the
computer circuits might be explored between tests and

provide important information on the potentialities

9 3
of the computer for simulator work.

In August the apparent need to make several changes

in the integrator circuits of the Analyzer represented

additional problems, while success with experimental

tests on a variable-oscillator design suggested

a feasible three-phase motor could be developed for

a variable-frequency servo application that the

Analyzer required. The project was makinp reasonable

progress on some details--Stephen Dodd was studying

the properties of the aerodynamic equations , George.

Schwartz was investigating ways and means of representing

aircraft piston-engine performances , and others

were examining aspects of radio noise level, cathode

follower characteristics, and analyzer component

interconnections — but progress with many of the

2.21

detailed design chores did not keep Forrester from
pondering upon the overall characteristics and
limitations of the simulator.

In retrospect it should be noted that the wartime
design experiences of the graduate students in the
Servomechanisms Laboratory, unusually rich and varied
though they had been, had not placed them at the fore-
front of innovative, analog- computer design activity
in the manner that their postwar experience in Project
Whirlwind was to give them pioneering competence and
pre-eminence in digital design work. Their relatively
heavy-handed, brute-force engineering approach to the
design of analog computation machinery contrasted with
the light touch manifested at the end of 1945 in the
analog computer approach taken, for example, by Arthur
Vance and his associates at RCA, Expert in designing
low-drift amplifiers, they developed driftless,
direct-current amplifiers that proved essential to
later analog computer development. Here they possessed
a degree of experience and competence that the
Servomechanisms Laboratory engineers lacked, and de
Florez' engineers in the Special Devices Division of
the Navy were aware of these differences.

The SDD program managers were also increasingly
preoccupied with the dawning missile and rocket tech-
nology that German engineers had launched spectacularly

2.22

with the V-2 rockets used to bombard London, and they
were sensitive to the greater challenges lying ahead

for simulator engineers in both the aircraft and the

24
missile fields of design. It would be easy to

suggest, in consequence, that the Navy programmers
deliberately began to encourage Forrester and his
colleagues to explore other design avenues that would
avoid the analog computer design problems they were
encountering , but. the evidence of such long-range
master planning is not only lacking but also contra-
dicted by the complex sequence and fortuitousness of
related events during the remainder of 1945. Ap-
parently aware that the course of engineering research
is not explicitly predictable since it requires inno-
vative intellectual activity if it: is to proceed, the
Navy engineers rested their confidence upon the already
demonstrated innovative abilities of the MIT engineers
and encouraged them to go whither their investigations
led them ,, within the overall: confines of the Airolane
Stability and Control Analyzer problem that had been
laid down.

Meanwhile, it was in August, 19.45.,.. while Forrester
and his: associates were seeking solutions to the analog
engineering problems bestting them, that Japan sur-
rendered.. The war was at last over, and many could
move once more to pick up the threads of their peace-
time lives- and occupations. Forrester had to eive some

2.23

of his attention to the necessary poinds and comings
and reorganization of activities that ensued, but
the dislocations of the war's end proved to be trans-
itory in their effect upon the project.

At the same time, in the wider technical community
and unknown to Forrester, a mathematician at Brown
University was making arrangements to call an inter-
national conference on computers in October. The end
of the war meant that a ireeting could be called to
take stock of wartime dvelopments, and R. C. Archibald,
chairman of a National Research Council committee, was
readying notices that would bring together experts
from England and the United States for a two-day
session at MIT. Archibald and his committee were
particularly interested in new "electronic devices. . .
which promise astronomical speeds for numerical
computing processes."

During the summer, Forrester had found time in
the midst of the routine of his laboratory affairs to
discuss computational techniques other than those
associated with the analog computer and found that
the engineering development of none was so far advanced
as to be of immediate use to him. From a fellow
graduate student in electrical engineering, Perry 0.
Crawford, Jr. , Forrester learned of the intriguing
future prospects that some already saw for employing

2.24

digital numerical techniques in machine calculation.
At MIT during the middle Thirties Professor Caldwell
had introduced a course in mathematical analysis by
mechanical methods, and Crawford, who had studied as
a graduate student and research associate under
Caldwell and Vannevar Bush, was well exposed to both
analogue and digital machine- computation concepts
when they were for the most part still in the con-
ceptual stage, especially where digital techniques were
concerned.

In 1942 Crawford had submitted his master's thesis
under the title, "Automatic Control by Arithmetical
Operations," setting forth one application of digital
techniques of computation, that of the automatic control
and direction of antiaircraft gunfire. After indicating
how recently physicists and electronic engineers had
become seriously interested in methods of performine:
arithmetical operations using such electronic
devices as the Eccles-Jordan flip-flop circuit, he
restricted his discussion to the problem of predicting
the future position of the target and described the
sort of electronic equipment that misht be built to
perform the operations required in automatic calculating:
"electronic switching elements, devices for multiplying
two numbers , finding a function of a variable, recording
numbers, translating mechanical displacements into

2.25

numerical data, and for translating numerical data
into mechanical displacements."^ Forrester recognized
that all of these operations were required of the
Aircraft Stability and Control Analyzer. There was no
question that Crawford understood the nature of his
problem, although it was not encouraging to hear
Crawford express the opinion that the successful
application of these new digital techniques to the
sort of problem that Forrester's group was attacking
lay still too far in the future to be of any help.

These ideas, presented to Forrester in stimulating

27

detail in mid-September, did not slip from his mind.

Meanwhile, there was the daily administration of the
project to attend to, and the momentum of project
affairs kept him busy. When Crawford left the Institute
a month later, to go to work for de Flore z in the
Special Devices Division of the Navy, he spent part
of his last day on the campus talking with Forrester
about digital calculators and the new breed of con-
trolled-sequence devices then coming over the horizon.
These were represented by an elaborate vacuum-tube
calculator, the "Electronic Numerical Integrator and
Computer," called ENIAC for short, and another known
as the "Electronic Digital Variable Automatic Computer,"

the EDVAC. Both of these were under development at

2 8

the University of Pennsylvania in Philadelphia.

2.26

While neither calculator was in operation yet , the
former was nearing completion at the Moore School of
Electrical Engineering under the direction of John
W. Mauchly and J. Presper Eckert, Jr. Mauchly was a
physicist, Eckert an electrical engineer, and both
were well aware that theirs was the first enterprise
committed to using vacuum-tube circuits to carry out
the complex calculations required.

Forrester was on the track of something new.
He liked what he saw, and the more he saw, the more
he wanted to see. He couldn't put it aside. In
after years both he and Everett were to attribute to
Perry Crawford the suggestion which they came to

take seriously, that digital numerical techniques

29

merited serxous study.

Although Forrester had worked more extensively
with analog devices , his mathematical and electrical
engineering background permitted him to recognize
and explore rapidly the prospects of digital calcu-
lation. There was nothing novel about the equivalence
of the two modes of calculating, analog and digital;
he was aware that these were alternative procedures ,
each possessing its particular virtues and defects.
And if the arrangement of some of the electrical
components that Crawford called to his attention was
novel, the tubes, capacitors, resistors, and elemental

2.27

circuits were familiar features that offered no
trouble. The novelty thus lay less in the elements
than in the system implications, and with these
Forrester promptly began to familiarize himself.
So intent were he and Crawford , and then Everett , in
their contemplation of the prospects , that they paid
little heed to the historical background of the state
of the art as they found it, and indeed such awareness
was not necessary to qualify them to carry out the
technical pursuit they then engaged in.

NOTES ON CHAPTER 2.

1 . Interview, Prof. G. S. Brown, MIT, by the author, July 6, 1964.

2. Interview, Jay W. Forrester and Robert R. Everett,
by the authors, July 31, 196 3.

3. J. W. Forrester, Computation Book No. 36 , entries of
Nov. 2, 1944.

4. J. W. Forrester, Computation Book No. 36 , entry of
Nov. 7, 1944.

5. J. R. Markham, 0. C. Koppen, J. Bicknell, Proposed
Method of Ensuring Satisfactory Handling Character -
istics in Airplanes , April, 1944; Specifications Tor
Airplane Stability and Control Analyzer , Navy Dept . ,
Bureau of Aeronautics, Special Devices Depot, N. Y. ,
N. Y., Aug. 11, 1944.

6. J. W. Forrester, Computation Book No. 36 , entry of
Nov. 7, 1944.

7. J. W. Forrester, Computation Book No. 36 , entry of
Nov. 21, 1944.

8. J. W. Forrester, "Hydraulic Servo mechanism Developments,"
pp. 4-5. This report was submitted "in partial fulfillment

of the requirements for the degree of Master of Science
from the Massachusetts Institute of Technology, Department
of Electrical Engineering, June, 1945."

9 . Ibid . , p. 6 .

10. Ibid . , pp. 6-7.

11. J. W. Forrester, Computation Book No. 36 , entry dated
Feb. 5, 1945, "Program to Date."

12. Servomechanism Laboratory Report No. R-100, Contract
No. 6345, March 4, 1945, subj . : "Method of Cockpit
Mounting and Actuating Mechanism."

13. Servomechanisms Laboratory, Project Whirlwind , Summary
Report No. 1 , (Apr., 1946), pp. 1-5; see also memo-
randum prepared but not submitted by J. W. Forrester,
subj.: "Status of Contract NOA(s)-5216 ," May 8, 1945.

14. J. W. Forrester, draft memo, 5-8-45, subj.: "Status
of Contract NOA(s) 5216."

15. Ibid .

16 . Ibid .

17 . Ibid . ; ltr. , N. McL. Sage, Director, DIC* to Chief,
BuAer, subj . : "Proposal for . . . Development for

. . .0. F. T.," May 22, 1945.

18. Interview, Everett and Forrester by the authors,
July 31, 196 3.

19. Ibid .

20. J. W. Forrester, Computation Book No. 39 , p. 34,
entry of May 26, 1945. r ~

21. F. L. Alt, Electronic Digital Computers (New York,
1958), p. IF!

22. Ltr., J. W. Forrester to Dr. G, S. Brown, June 22,
1945.

23. J. W. Forrester, Computation^ Book No. 39 , supplement
to p. 54, entry of June 27, 1945. "

24. Interview, P. 0. Crawford by the authors , Oct. 25,
1967.

25. Ibid .

26. P. 0. Crawford, Jr., "Automatic Control by Arithmetical
Operations," "submitted In partial fulfillment', of the
requirements for the Degree of Master of Science at
the Massachusetts Institute of Technology, 1942,"
pp. 1-2.

27. J. W. Forrester, Computation Book No. 39 , suppl. to
p. 65, entry of Sept. 18, 1945.

28. Ibid . , p. 42, entry of Oct. 16, 1945.

29. Interview, J. W. Forrester and R. R. Everett, by the
authors, July 31, 196 3.

THE SHIFT TO DIGTAL

The summer and fall of 1945 found a small but growing
interest in electronic digital computers flourishing here and
there in the United States and Europe. It was of little conse-
quence that no such electronic digital computers were yet in
operation, so far as their attractive potentialities were
concerned. What mattered was that western mathematicians
and engineers were beginning to be caught up in a classic
example of the historical phenomenon of "convergence, "
in which the embryonic computer technology was assuming
its shape and character from the joining together of several
diverse machine design traditions and several abstract
intellectual traditions. Personal curiosity had combined
with historical circumstance to place various individuals
at peculiar, strategic positions from which they could take
advantage of the opportunities provided by this convergence
of traditions. Among these individuals happened to be the
young engineers at the Massachusetts Institute of Technology.
Their individual exertions had helped to bring them to such
positions, and the converging traditions set the boundaries
within which their ingenuity would go to work.

3. 1

3.2

The first of these traditions was itself a composite of
several machine and device developments, each of which had
a long history. They included the odometer and the abacus in
diverse forms from ancient times, the slide -rule and the
mechanical adder from recent times, and the electromechanical
calculator traditions of the last hundred years, culminating in
the relay machines of George R. Stibitz and his colleagues at
the Bell Telephone Laboratories and the electromechanical
"Harvard Mark I" built by Howard Aiken with the assistance
of several associates from the International Business Machines
corporation.

The intellectual traditions included at least three or four
of note -- those of counting and "reckoning, " or calculating;
other mathematical traditions that had produced the logarithm,
the slide-rule, and Charles Babbage's unbuilt and forgotten
computer (to name only a few): and the scientific and technical
traditions that produced, in one direction, theories about
electrons and electromagnetic phenomena, and in the other
direction, applications in such forms as the vacuum tube,
electronic circuits, and radar.

The technical machine tradition that first exploited
computer techniques in the late 1930s and early 1940s was the
electromechanical tradition brought into being by prior advances
in the technology of electricity, in the technology of controlled-

3.3

motion machinery, in the technology of the desk calculator,
and in the technology of the developed telephone system. At
the hands of Aiken, Stibitz, and others, it provided the first
machine generation of computers, but although it spawned
refined and improved members of one class of relay, switch,
and gear machines, it shortly proved to be a subclass of
provocative but sterile computers without lasting issue. The
cause lay not in any essential "hybrid sterility, " but in the
development of a more promising machine form that happened
to be just a trifle slower to achieve practical working condition.
This second form permitted vastly greater calculating speeds,
and it provided a second machine generation of computers
mechanically unrelated to the first: the electronic digital
computers.

Electronic digital computers came hard on the heels
of Aiken's and Stibitz's innovations. The first true electronic
computer was the Electronic Numerical Integrator and Computer,
better known by its acronym, ENIAC, designed and built under
the direction of John W. Mauchly and J. Presper Eckert, Jr. ,
at the University of Pennsylvania for the United States Army.
Mauchly as a physicist had found himself (without realizing it)
in the same predicament in 1941 that Aiken and then Stibitz

3.4

independently suffered: he was sharply aware that great
quantities of data needed mathematical processing and that
existing techniques were slow, cumbersome, unreliable, and
hopelessly unable to meet the challenge. He concluded that
electronic facilities ought to be exploited but found no one in
electronics at work on the problem. So in the fall of 1941 he
joined the Moore School of Electrical Engineering at the
University of Pennsylvania, determined to obtain the electronics
design assistance he sought.

By 1943 he, Eckert, and others at the Moore School,
as well as Army Ordinance representatives, were convinced
that a system of vacuum tubes and radio circuits, of standard
form and established characteristics wherever possible, should
be able to perform lengthy digital computations in the laboratory
with the necessary speed and accuracy to provide values
extremely useful for further perfecting greatly needed ballistics
trajectory data. Their ensuing work on the ENIAC during 1944
and 1945 was carried on under wartime secrecy, but this did
not keep it from coming to the attention of the Princeton
mathematician John von Neumann, or of some of the electrical
engineers at MIT. The machine went into operation in 1946
and was put to work on ballistics calculations as planned.

3. 5

It would multiply two ten- decimal numbers in less than three
thousandths of a second, compared to the three- second interval
the Mark I required.

The ENIAC was deliberately designed to be of limited
use and was not intended to be a general-purpose machine of
wide application, for there were many engineering problems
that its designers had to solve without ambitiously extending
the capacities of their projected machine. In this respect they
resisted the temptations to enlarge and improve that had caused
Babbage to fail a century earlier. Aiken's non-electronic Mark I
filled a wall, so to speak, fifty- one feet long and eight feet high,
and Bell Aberdeen relay computer was housed in a room approxi-
mately forty by thirty feet in area. The ENIAC also was large,
occupying the walls of a room approximately forth feet by twenty
feet in size and including racks of assemblies on wheels in
the center of the room. 18,000 radio tubes and 1, 500 electrical
relays went into its construction, together with plug boards,
wiring, power units, and related equipment. Its internal
storage was kept small, consonant with the mathematical
chores it was expected to perform. The large size of this
and other early computers was deliberately stipulated by
the designers. Access to the components was what they

3.6

were interested in, so that unanticipated repairs and
improvements might be made easily in these pioneer
machines.

The ENIAC, being electronic, was not an engineer-
ing descendant of the Mark I mechanical machine. Instead,
its physical equipment tradition traced back into the complex
history of electronics in radio and telephony. The electronic
computer tradition after the ENIAC rested not only on the
ENIAC itself but on wartime developments in the pulsed
circuitry of radar, on the well developed state of the art
in radio tubes and circuitry, which was subsequently modified
by transistors and solid-state circuitry, and on the logical
abstractions of a mathematical tradition that included such
names as Babbage, Edward Boole, and John von Neumann.

As had been the case with the abacus tradition and
the mechanical calculator tradition, so it was with the
electronic computer tradition, for this tradition arose out
of the accumulation and synthesis of many strands of
endeavor associated previously, as well as then and later,
with other traditions. Contributing to this synthesis was
the abstract logical, but not the equipment, tradition of
the automatic sequence controlled calculators. The proto-
types of Babbage, Stibitz, and Aiken, the first premature,
the latter two within the practical state-of-the-art, had

3.7

come into existence as evolutionary consequencies of a
long and complex historical tradition of their own. When
this calculator tradition was joined to the hitherto separate
electronic traditions of radio and radar circuitry, their
coming together seemed so natural, and the modes of
application of circuits to computation and of computational
and logical concepts to electronics were so provocative,
that the specialists involved did not even wait for the first
electronic computer, the ENIAC, to be completed, tested,
and put into operation before they rushed on enthusiastically
to more ambitious designs.

It was at this juncture of events that the MIT engineers
working on the Aircraft Stability and Control Analyzer fell
in with those who were engaging in the activities bringing
together the calculator and the electronic traditions. It
was Forrester's good fortune, as the leader of the group,
to find himself standing at the intersection of the two
traditions; his was also an instance of the innovative
situation that Pasteur once characterized with the remark,
"Chance favors the prepared mind. " Forrester's mind
was indeed prepared by the computer problems the aircraft

3.8

analyzer was posing at the time, and he grasped at this
attractive prospect of a general solution.

It was also Everett's and Forrester's combined
insight and vision to perceive, seize, and exploit the
opportunities of joining and applying the abstract,
elaborate, logical- formal tradition of manipulating
discrete numerical quantities (upon which the mechanical-
calculator tradition has been resting) to the engineering
tradition of producing radio and radar equipment and, in
the course of their enterprise, to expand and develop
further both of these traditions. Their intuitive and
analytical assessments of the rate at which such reduction
to practice (for some reason it is never called "elevation
to practice") could be accomplished was one key factor
that was to make their particular computer unique in its
time. A second key factor was the type of task they
designed their computer to perform. In this respect their
computer was so unlike all the others that for a while it
appeared it would find no use to justify either its cost or
its very existence. Then once again a separate tradition
over which Forrester and his group had had neither

3.9

control nor influence - - this time taking the form of
certain highly specialized preparations for waging war --
joined with and made use of the brand-new electronic
computer tradition.

But that is getting ahead of the story. At the end
of summer, 1945, the prospect confronting the MIT engineers
was still that of a cockpit or control cabin connected, some-
how, to an analog computer related to the design tradition
of the electromechanical differential analyzer developed by
Vanner Bush and Samuel Caldwell between 1935 and 1942.

News of the successful operation of the MIT differential
analyzer had been withheld during the war lest the enemy
learn how useful and practical it was. Instead, intimations
were deliberately "leaked" that it had failed to live up to
expectations. The end of hostilities permit MIT to celebrate
the true achievements of the analyzer at a meeting of computer
experts which has already been mentioned -- that which
Professor R. C. Archibald of Brown University had called
in the name of the National Research Council, acting as
chairman of the Council's Committee on Mathematical Tables
and Other Aids to Computation.

3. 10

The meeting took place on October 30-31, 1945
at MIT under the auspices of "Subcommittee Z on Calculating
Machines and Mechanical Computation, " and the purpose was

expressed in the formal title: "Conference on Advanced

2
Computation Techniques. " Among those who attended were

Perry Crawford and Jay Forrester. The latter was particuarly

interested in reports of the design activity going on at the

University of Pennsylvania. Prepared by his conversations

with Perry Crawford and anyone else he had encountered

who was knowledgeable, he was keenly receptive to the

stated object of the Conference "to familiarize each member

of the Group with present potentialities in the field, and to

make known future developments. " The program of papers

to be delivered and especially the roster of expected conferees

redoubled both his curiosity and his growing commitment

to the new digital tradition then emerging.

Before two weeks had passed, he had visited the

University of Pennsylvania to obtain more information and

was inquiring into the design details of the ENIAC and its

projected successor, the Electronic Digital Variable Automatic

Computer (EDVAC).

3. 11

It is difficult, if not impossible, to say confidently-
just when the realization struck Forrester that a novel
solution -- the electronic digital instead of the analog mode --
had presented itself. Caldwell's suggestion in May that
Forrester might find Stibitz's work significant did not
precipitate the vital moment. Did the conversation with
Perry Crawford in September, then? Or was it their
discussion on the day Crawford left in October? In after
years, Forrester recalled standing on the steps in front

of one of the Institute buildings talking with Crawford, when

3
the latter' s remarks turned on a light in his mind. His

recollection is that from that time on, he began to consider
the digital mode seriously.

Such authentic evidence must be used with caution,
nevertheless, for historical analyses of inventive activity
have shown time after time that our use of hindsight to
reconstruct and evaluate crucial events responds to the
logical and esthetic requirements of making sure that what
went before fits rationally with what is later known to have
followed. However satisfying and consistent such reconstructions
are, even though based on first-hand knowledge and participa-
tion, they ignore this historical fact: that which came before,

3. 12

was brought to pass without the knowledge of after events
that hindsight subsequently bestows. Whatever the personal
theory of the inventive process the innovator himself holds,
and whether this theory has been rigorously thought through
and made explicit in his own mind or whether it is rough- cut
or assumed intuitively without demonstration, this is the
innovator's private theory of the historical process --
although it is seldom regarded as the historical process,
under such circumstances -- that enables him to reconstruct
the past to his own satisfaction.

The intent of these remarks is not, of course, to cast
doubt upon the value or honesty of personal recollections but
rather to point out how very complex and how independent of
both rationality and irrationality is the process by which new
things occur. At the heart of the problem of historical
reconstruction of events lies the grave risk of generating
misunderstanding and confusion by unwittingly co-identifying
our assumptions, and consequently our views, regarding the
process by which new things occur, with the process itself.
Experience demonstrates that most innovators have been
too busy in their own fields of thought and action to work
out a rigorous analysis of how events proceed. They consider

3. 13

it a waste of time, if not an affront to their integrity, to
consider how they know what they very well remember, or
how well their recollections accord with prior and subsequent
events in detail. Such considerations remain the professional
responsibility of the historian as he seeks to approximate
and interpret the past.

In any case, the turn of events that strategically
affected Forrester's prosecution of the ASCA project is
conspicuously visible in internal developments within the
project, in the hiring practices and the efforts to secure
the assistance of certain departments within the Institute,
in the renegotiation of the Navy contract, and in Forrester's
and Crawford's separate and joint efforts to explore the
prospective application to other uses of a digital computer
sophisticated enough to meet the requirements of the aircraft
analyzer.

The months of November and December following
Archibald's computer conference were put to use by
Forrester in testing the immediate import and implications
of the digital concept. It was a period during which he
critically tested, revised, examined, and re-examined the
value of the insight he had had regarding a solution to his

3. 14

problem. If that insight were glitter without substance or

too wild a dream, then it must be dismissed. If it continued

to show promise, the grounds for this promise must be laid

out and at least prospected in a preliminary way.

Accordingly, on November 9 Forrester, as part of

his effort to shift the attention of his staff, held a conference

of his own project engineers on the general subject of

techniques in computation. The opening paragraph of the

report of this conference, written by one of the project

engineers, reveals how fundamental was the reexamination

and the reorientation that Forrester was initiating:

Analogy type of computation has been
under consideration, but there are other
approaches which are highly thought of in
mathematical circles throughout the country
at present. It is the purpose of the present
discussion to consider certain features of
some of these.

After indicating the basic modes of calculation of
the Harvard Mark I and noting that its mechanical
calculating speeds were quite slow, the report turned
to the "Pennsylvania technique" of calculating electronically
with vacuum tubes. "Even though the machine they are
building along these lines has not worked yet, there is

3. 15

already a proposal to make another one which contains other

features of importance to the aircraft analyzer problem, "

5
said the report. The latter machine referred to was the

EDVAC, which was intended to employ pulses in temporal,

linear sequence and store its information "by the use of a

thin column of mercury, referred to as a 'tank' 11 --a

mercury delay line.

A possible method of adding was next discussed,
then a principle of operation of multiplying that would take
only two or three times as long as a single addition. The
report's closing remarks reflect how sharply Forrester and
some of his engineers were examining the novel digital design
trends: "These digital techniques are fundamentally processes
of doing one thing at a time. Techniques for high speed
electronic computation have not been worked out, and
several months' work will be necessary to properly evaluate
the process. " Actually, the evaluation was already under way
in the Servome onanisms Laboratory.

On November 13th, after visiting the University of
Pennsylvania, Forrester requested from the Pentagon access
to published technical reports on the EDVAC, pointing out that

3.16

his Navy aircraft research contract involved "the simult-
aneous solution of many differential equations, and the
techniques visualized by the University of Pennsylvania on
their EDVAC computer show considerable promise. "

By the middle of November he was looking for
additional specially trained and talented young men to
add to the project staff, men who could move with the old
staff in the new direction they were heading. He sought
men who had records of special competence as excellent
graduate students or as well- qualified radar experts.
Lip-service to the policy of top- quality work from top-
quality personnel would not suffice; he would consider only
men of exceptional promise and not waste his time on the
run of the mill. Thus, in a letter of November 16th to the
MIT Radiation Laboratory he asked for men with a radar
design background, "experienced in the generation and
handling of low -power level signal pulses, in applications

which will require pulses a part of the microsecond long

7
and spaced a microsecond apart. " Suitable prospects

ought to be between 25 and 40 years old, he specified,

"and of doctor's degree caliber although selection will

be based on the man's experience, cleverness, ingenuity,

3. 17

and references rather than on his academic degrees. "
He closed the letter on a characteristic note: "I wish
to stress the need for cleverness and ingenuity in the field
of coded pulses. "

He was convinced, from his wartime experience,
that it was more efficient to pay a higher price for one
really good man and then give him his head, than it was
to pay less for three average men and have to lead them
by the hand. He felt he possessed the experience and the
judgment to enable him to recognize the difference, and
although he did not expect his every selection of a new
man to be error-free, he was adamant about maintaining
exceptional standards in choosing design and test engineers
and mathematicians. Over the years ahead this policy
brought in many promising men, some of whom went on to
important industrial, engineering, and consultative
achievements in their subsequent careers.

The high standards that Forrester and his associates
came to insist upon, with the endorsement of Brown and
Sage in the background, inevitably attracted unfavorable
comment from some outsiders as time passed, eliciting

3. li

off- the- record remarks that the project personnel were
arrogantly high-hat and snobbish, working in a building
closed by security regulations to outsiders, that they were
as unrealistic about what they were doing as they were
young and immature, and that theirs was a "gold-plated
boondoggle, " extravagant in its demands, in its rewards,
and in its raids upon the taxpayers' purse.

Unquestionably, it became for the project members
a deliberate policy of saving development time and money
in the long run by insisting on going "first class. "
Drawing together as it did young men of ambition,
ability and spirit, and reinforced by a habit of daily
operations that stressed and, for the most part, obtained
intelligently planned and coordinated operations, this
policy produced an unusually high esprit de corps.
As one project member recalled years later, viewing the
operation from the perspective of a personally successful
administrative engineering career in the computer hard-
ware business, "We were cocky. Oh, we were cocky!
We were going to show everybody! And we did. But we

3.19

had to lose some of the cockiness in the sweat it took to
pull it off. "

Forrester's philosophy of acquiring high-caliber
staff members became the project's continuing policy,
and representative events that helped bring this about
are to be seen in measures he took to enlist the aid of
certain departments at the Institute, even as he was
beginning to explore in earnest the possibility of using
the still thoroughly "new-fangled" digital computer, a
successful vacuum-tube model of which had never yet
been put into operation nor any other electronic model,
for that matter. A week before Christmas he was suggesting
active consultative arrangements. "I feel," he wrote to
Professor Henry B. Phillips, "that the Mathematics
Department can make an outstanding contribution to our
work in studying proper set-up procedures and techniques
to be used in digital computation, " and he suggested that

exploratory conferences with specified individuals be

o
arranged to generate momentum. Already, since early

December, lecture notes on mathematical analysis by

mechanical methods were being provided to ASCA personnel

3.20

from a course given by Professor Samuel H. Caldwell of

9
the Electrical Engineering Department.

Forrester's letter to the Physics Department was

more detailed than his letter to Professor Phillips, and

it revealed his awareness of the prowess a digital computer

would possess if it could meet the requirements of the

Aircraft Stability and Control Analyzer: "It is our hope

that the solution of the equations involved can be accomplished

by electronic techniques. If so, the computer also will be

capable of solving many other problems in the fields of

mathematics and physics. " Such a computer, he recognized,

would be "of much greater scope than any other now in existence

or being considered for the immediate future. " He hoped

that they might be able to obtain the "active participation of

certain men whose primary interest is in physics, but with

a secondary interest in our work. " He named a couple of

graduate students then matriculating for the doctorate as the

sort of talent he was looking for, and he indicated that suitable

appointments could be made either through the Electrical

Engineering Department (with which the Servomechanisms

Laboratory was affiliated) or through the Physics Department,

as they preferred.

3.21

At that time he thought of the program in prospect
as comprising two major parts: a six-month "preliminary
survey of electronic computation possibilities, " and a
three-year design and construction phase. This view,

which had come out of discussions and personal reflections
before the end of November, sifted out the following
considerations: If the analogy method of solution were
pursued, using a physical representation of the problem
and measuring the physical quantities of interest, and if,
to this end, the Differential Analyzer type of analog machine
were used, it would unfortunately not be suitable for the ASCA
problem. To represent the quantities by electrically inter-
connected mechanical shaft rotations, for example, would
require a device of extreme complexity. Further, the length
of time to accomplish a solution would be prohibitively

impractical, "because equipment will not respond fast

12
enough to give the pilot proper 'feel'. "

On the other hand, if ASCA computation were performed

by electrical voltage analogy instead of mechanical shaft

rotation, as Forrester and his assistants had been thinking

3. 22

of doing, then one encountered the technical difficulties of
insufficient signal range. These along with the factors of
the physical tolerances which could be achieved and of the
friction which would be encountered, might severely limit
sensitivity and accuracy of solution. A change in equations
or problems would require elaborate new physical hook-up
of the operating components. However, analog solutions
ought to be exact, in theory, and the analogy technique was
well known and well established for relatively simple problems.
Such were the pros and cons on the analogy side of the
question.

On the digital side, numerical analysis by arithmetical
processes would replace analogous physical quantities and
could solve the entire set of ASCA equations, given time
enough. If there were not time enough, then there would
have to be speed enough. There would be no physical
tolerances or friction to contend with, and although digital
techniques were not so well established as analog, experience
was being acquired with the Harvard Mark I, the Bell Telephone
Relay Computer, and the University of Pennsylvania ENIAC
then approaching completion. "Most new plans, " Forrester
observed, "lean toward the binary system" of notation, and

3.23

the vacuum tube, he knew, was a more reliable -- and

13
incredibly faster -- binary (on- off) device than a relay.

Considering the prospects of a hypothetical electronic machine,

then, one might expect two types of storage, mercury delay

line storage and electrostatic tube storage, and suitable

computation and control circuitry.

He saw various mathematical and engineering

advantages in digital computation, not the least of which were

the prospects that construction costs might be less and

trouble location easier. Problems could be set up more

rapidly and since their solution must progress one step at

a time in the digital mode, then "the problems of a large,

interconnected, simultaneously operating analog computer

network are avoided. " Finally, the computer could be "used

14
for many problems other than aircraft analysis. "

Nevertheless, the digital disadvantages were formidable.

Although construction costs might be lower, development costs

might be higher, for digital techniques and devices were not

well known or well established. Development would take

more time, but construction should take less. Further, he

reflected realistically, the ASCA problem "requires pushing

the digital technique well beyond anything even contemplated

15
up to the present time. "

3.24

Although many circuits appeared to be ready for
development and use, five or six months of intensive study
would be necessary in order to find whether the first
promise were as substantial as it appeared. Forrester
concluded that such a study must be made. The prospects
were sufficiently attractive and enough people were
optimistic about the future of digital computation to
warrant proceeding a step further. Was this a prudent
decision? It certainly was not without risk. "This group

at the present time, " he wrote, "has no concrete informa-

1 6
tion on which to predict the outcome of an investigation. "

By mid- January 1946, the investigations which

Forrester and his associates had been carrying on since

October gave Forrester sufficient confidence to recommend

to the Navy that "numerical electronic methods as applied

to the aircraft analyzer be carefully investigated. " If the

digital computer could be successfully developed, their

proposal noted, the rewards would be great; among them

would be "more reliable performance, higher accuracy,

lower cost, smaller size, and more flexible operation. "

3.25

In addition, the digital computer would permit the "solution

of many scientific and engineering problems other than those

17
associated with aircraft flight. " This last comment

contained within itself the embryo of the general multi-
purpose computer which was increasingly to become the
primary goal of the project.

Digital computation techniques and methods, however,
were not accepted immediately as a panacea. Extensive
research was necessary, with all the ramifications such a
program entailed, but the advantages appeared so attractive
that the Institute proposed that the new method be thoroughly
explored. To this end a contract was proposed which sought
the accomplishment of two tasks: (1) the design of a digital
type computer adequate to the requirements of the aircraft
analyzer; (2) the adaptation of the computer to the analyzer,
and the design of required associate equipment. The two
tasks would overlap chronologically. Task One would
commence immediately and pursue intermediate objectives
until December 1949, when the whole project was to be
completed. Task Two would phase in around December 1946,
but only after a "final decision on the practicability of

3.26

electronic computation as related to the aircraft analyzer''
had been made. Task One, it was estimated, would cost

$1. 910, 000; task two $477, 600 -- a total of $2, 388, 000

1 8
for the completed project. Such was their thinking at

the beginning of 1946. If the projected rate of progress

was overly optimistic, it nevertheless provided timed

goals to aim for.

Simultaneously with the submission of the proposal

to the Office of Reserach and Inventions, Forrester replied

to a request made earlier by Lieutenant Commander

H. C. Knutson of the Special Devices Division of the Office

for "comments on the applications of high-speed electronic

19
computation. " Forrester's reply contained much of the

substance of ideas that had emerged from previous discussions

he had held with Perry Crawford on the subject. The digital

computer, Forrester predicted would possess a flexibility

not possible with the analogue computer, permitting therefore,

the construction of a "Universal Computer" with definite

possibilities for military application in both tactics and

research. In tactical use, it would replace the analogue

computer then used in "offensive and defensive fire control"

3.27

systems, and furthermore, it would make possible a
"coordinated CIC (Combat Information Center), " possessing
"automatic defensive" capabilities, an essential factor in
"rocket and guided missile warfare. " In military research,
electronic computation mcide possible wide and diversified
research programs in "dynamic systems": (1) aircraft
stability and control; (2) automatic radar tracking and
fire control; (3) stability and trajectories of guided
missiles; (4) study of aerial and submarine torpedoes
including launching characteristics; (5) servomechanisms
systems; and (6) stability and control characteristics of
surface ships. Digital computation, furthermore, would
allow the "study of both interior and exterior ballistics"
and "stress and deflection studies in ship and aircraft
structures. "

Leaving the areas of possible military application,
Forrester turned to a detailed analysis of the implication
of high-speed electronic computation for scientific and
technical research in general. Here he predicted wide-
spread opportunities in the fields of (1) nuclear physics,
(2) thermodynamics, (3) compressible fluid mechanics,

3.28

(4) electrodynamics, (5) mechanical engineering, and
(6) civil engineering. He further considered its applica-
tion to statistical studies in both the physical and social
sciences. In the latter sciences alone, he observed, it
would be of value to government agencies and departments.
He concluded his response with the following comment:

The development of electronic digital
computation is only beginning, and consider-
able effort and money will be expended in
achieving the equipment to meet the above
objectives. Once sufficient development is
completed, however, the cost of duplicating
electronic computing equipment will be less
than for other forms of computers. Beginning
with a suitable basic design, new computers
could be built with facilities for a specific
magnitude of problem by adding or omitting
standardized memory or storage units with-
out requiring significant redesign.

The proposal that had been made in January to the

Chief of the Office of Research and Inventions was resubmitted

the following March in revised form. Substantively, the

revision differed little from the original: it requested that

the date of completion be extended to June 1950 and that

the total allowable expenditures be increased from $2, 388,000

to $2, 434, 000. In addition, a summary of Forrester's

letter to Knutson was embodied in the revised proposal.

3.29

Four principal tasks were delineated by the Institute:

1. "Research, development and contrauction
necessary to demonstrate digital techniques of the type
required for the final computer. "

2. "Design of a computer which is adequate
for the aircraft analyzer problem. "

3. "Construct and assemble the computer and
associated equipment for control and stability studies on
aircraft. "

4. "Operation of the complete equipment for

the solution of aircraft stability problems and application

21
of the computer to other types of scientific computation. "

Although the revised proposal submitted to the Office

of Research and Inventions was an Institute document, it

also reflected the influence and ideas of those at the Special

Devices Division who were immediately responsible for Navy

administration of the trainer -analyzer project. The rapport

between the two groups was sufficient to produce fruitful

joint discussions in which a proposal acceptable to the

Navy could be worked out. Actually, the Special Devices

3. 30

Division in February 1946 had recommended to its parent
organization, ORI, that the Institute's proposal be accepted
and implemented. Hence, it is probable that the March
revision reflected from the Office of Research and Inventions
on the original proposal. The revision was then transmitted
to the Institute by the Special Devices Division. As a
consequence of whatever internal adjustments were made,
the Office of Research and Inventions incorporated Tasks
One and Two of the March proposal into Contract N5ori-60.
This contract superseded the earlier Letter of Intent for
former Contract Noa(s)-7082 and became retroactively
effective to June 30, 1945.

Under Task Order I of the new contract, the Institute
was to undertake first the construction of "a small digital
computer involving investigation of electric circuits, video
amplifiers, electrostatic storage tubes, electronic switching
and mathematical studies of digital computation and the
adaptation of problems to this method of solution. " Second,
it was "to design an electronic computer and aircraft analyzer
based on Phase 1 of this Task Order. "

3. 31

Phase One was to commence as of the date of the
contract and was to terminate on June 30, 1947. The
second phase, commencing on July 1, 1947, was to
terminate on June 30, 1948. The total cost of the contract

was set at $1, 194, 420; the first phase would require $666, 000,

22

the second the balance of $528, 360. These costs were in

agreement with the amounts set by the Institute in its revised
proposal of March 1946. In that month, also, the Navy
revised its initial specifications to conform to the changed
conditions and goals. Finally, in the revised specifications,

the project was given the name by which its was to be known

23
in the future: "Whirlwind. "

NOTES TO CHAPTER 3

1. V. Bush, S. H. Caldwell, "A New Type of Differential
Analyzer, " Journal of the Franklin Institute, Vol. 240,
no. 4 (October 1945), pp. 255-326; S. H. Caldwell,
"Educated Machinery, " Technology Review , Vol. 48,
no. 1 (November 1945), pp. 31-34.

2. Conference program, October 30-31, 1945, entitled:
"Conference on Advanced Computation Techniques,
National Research Council, Committee on Mathe-
matical Tables and Other Aids to Computation,
Subcommittee Z on Calculating Machines and Mech-
anical Computation; Cambridge, Massachusetts. "

3. Interview, Jay W. Forrester, by the authors,
July 31, 1963.

4. Conference Note C6, written by Kenneth Tuttle,
Subject: "Conference on Techniques of Computation
held November 9, 1945, " November 14, 1945.

5. Ibid, p 3.

6. Letter, airmail special delivery, J. W. Forrester
to Office of Chief of Ordnance, November 13, 1945.

7. Letter, J. W. Forrester to James W. Walsh,
November 16, 1945.

8. Letter, J. W. Forrester to Prof. Henry B. Phillips,
December 17, 1945.

9. Eng. Memo No. 3, Subject: "Lecture Notes - Part I . .
December 5, 1945.

10. Letter, J. W. Forrester to Prof. John G. Slater,

December 17, 1945. A carbon of each of these letters
went to N. Mel Sage and Gordon S. Brown for their
information.

NOTES TO CHAPTER 3 (CONTINUED)

11. Ibid.

12. Conference Note C9, written by J. W. Forrester,
Subject: "Outline of Discussion on Digital Computation

as Applied to the Aircraft Analyzer, " November 28, 1945.

13. Ibid, p. 5

14. Ibid, p. 8

15. Ibid, p. 9

16. Ibid.

17. Ltr., N. McL, Sage to Chief, Research and Inventions,
USN, January 16, 1946; see also ltr. , A. P. Bencks,
Lt. USNR. SDD (Washington), to N. Sage, DIC, MIT,
November 27, 1945.

18. Ltr., N. McL. Sage to Chief, Research and Inventions,
USN, January 16, 1946.

19. Ltr., J. W. Forrester to Lt. Cmdr. H. C. Knutson,
SDD, ORI ("Washington), January 28, 1946.

20. Ibid.

21. Memo, J. W.. Forrester to N. McL. Sage, February 25, 1946;
memo, N. McL. Sage to J. W. Forrester, March 16, 1946.

22. Navy Department, Office of Research and Inventions,
Contract Number N5 ori-60, June 30, 1945, "Task
Order I -- Constituting a part of Contract N5ori-60
with the Massachusetts Institute of Technology and
superseding BuAer Letter of Intent for Contract NOa(s)
7082, " (The documents do not explain the apparent
$60 discrepancy between the two totals. )

23. Navy Department, Office of Research and Inventions,
Special Devices Division, "Specifications for Project
RF-12 known as WHIRLWIND, " revised March 20, 1946.

'U^i J- uux

PRELIMINARY DESIGN EFFORTS

At the end of 1945 the computer was still the tail
of the dog, so to speak, and the Aircraft Stability and
Control Analyzer was the dog. A year later, judging
by events within the Servome onanisms Laboratory, the
tail had passed through and beyond the point of wagging
the dog and had become the dog. The Analyzer became
the tail, and even that was cropped in 1948 when the
cockpit was junked.

It could be argued in after years that abandonment
of work on the rest of the Analyzer was a serious tactical
mistake, for as the Analyzer faded further into the back-
ground, so did the once-obvious immediate practical
relevance of the untried computer become more remote
and more nebulous. The overriding pragmatic question
in the spending of military funds was, classically, "What's
it good for? " Although the answer became dazzlingly
clear to Forrester and his associates in the project and

4.1

4.2

was perhaps earlier as clear to Perry Crawford, it
seemed to become less clear to outsiders almost in
proportion to the rising costs, as the months and years
passed.

Early in 1946 Forrester at the Massachusetts
Institute of Technology and Crawford in the Navy separ-
ately saw five years of research and development work
ahead before demonstrated success would be perceived
and appreciated. To Forrester it became a goal that
required the sort of monthly pace a $100, 000-a-month
budget would provide.

At the end of the first week in January he was press-
ing his search for the men he needed. "The general type
of man whom we need, " he wrote in a letter asking Nat Sage
for assistance, "should have originality and what is often
referred to as 'genius. ' He should not be bound by the

traditional approach. ... I do not know of suitable

1
prospects. ..."

While he was beating the bushes for top personnel,

he set up ten divisions in the Laboratory - seven to carry

on the technical work and three to support these -,

4.3

ordered a weekly meeting of each division at a time
when he could be present, created a coordinating
committee of his divisional leaders, and called for a
stepped-up delivery of reports on technical progress
and problems. "The Navy expects, and rightly so, to
be informed of research and development progress

through suitable reports, " he pointed out to his staff

2
in an early "Conference note. "

By the end of February he had a firm enough pro-
spective schedule, based on discussions with the Navy,
to call several tasks and their time schedules to the

attention of the engineers on the project. These tasks

3

covered the time period from July 1945 to June 1950.

According to the new schedules, the last six months of
1945 had been devoted to completion of studies in analog
computation, to preliminary investigation of digital
electronic methods, and to plans for carrying on the
latter. Characterized in this manner, the project gave
the impression of being routinely in command of its
situation at all times, and this was an impression that
Forrester sought naturally and by design to convey to his

4.4

But had they been in command of the situation at all
times? This was a question which Navy programmers and
administrators were to raise later more than once, not
only regarding this particular period in the affairs of the
project but regarding later periods as well. There were
some critics who came to feel that Forrester was attempt-
ing to gloss over the brute fact that the project had had to
abandon its first intention to build an analog computer, just
as, later, it abandoned the Aircraft Analyzer. Forrester
and his associates had made a false start, ran this argument.
What was to be gained, then, beyond self-deception and false
impressions conveyed, by describing the situation otherwise?
And why try to deceive his own engineers, many of whom had
been intimately involved, by such statements in his published
schedule?

The answer is to be found in part in Forrester's style
of conducting his affairs and in part in the character of the
research-and-development process. He saw himself as
best carrying out his directorial function by shielding his
men from potential outside interference that would inter-
rupt their progress and, at worst, demoralize their

4.5

enterprise. It was his responsibility to see that the
project had what it needed to proceed with its investi-
gations and to not distract the efforts required to
proceed by making the personnel of the project privy to
external administrative, policy and fiscal problems
that they were not qualified to handle, that they were
not hired to handle, and that they could do nothing about
in any event. Forrester saw no reason to allocate time
for his engineers to stand wringing their hands.

Since both he and his staff understood the difference
between the known and the unknown and between the
predictable and the unpredictable in engineering research,
no false illusions were being generated within the organi-
zation by putting the best face on the fact that preliminary
views and preliminary investigation had yielded unfore-
seen negative information that stimulated the discovery
of an affirmative alternative. As Forrester well knew,
an engineering problem was also an engineering opportunity
the validity of which could be affirmed only by the finding
of a solution. The digital computer offered a challenging
and exciting solution, indeed. So he chose to regard the

4.6

last six months of 1945 as a period devloted to analogue
and digital computation studies rather than as a period
of crisis, and although it had been a time when far-
reaching decisions were made, these did not constitute

...... 4

a serious crisis, m his view.

He gave his team a year to lay its plans for building
a digital computer, another half year on simulator cockpit
studies and equipment and on logical designs and bench-
test models, a third year to work up final equipment and
circuit designs and to begin work on final components, a
fourth year to build the prototype computer and receive
the cockpit from the Navy as an item of "Government-
Furnished Equipment, " and a fifth year to finish, test,
and deliver to the Navy the completed analyzer. In the
middle of 1950 the machine would go into full operation.

He visualized four tasks. The first would produce
a small digital computer that could perform the basic
functions and would see the accomplishment of basic
theoretical work by the middle of 1947. The second
would begin before Task I ended, would last for a year,
and would lay out the basic designs for the cockpit and

4.7

prototype computer. The third task would overlap the
second and in a year produce these components of the
prototype Aircraft Analyzer, and the fourth task would
produce the working, tested Analyzer a year later.
In conclusion, Forrester stipulated a policy of
periodic review "because of the indefinite nature of
the problem and dependence upon ideas which have not
yet been formulated. " He also made it clear that the
schedules and tasks described were not fixed for all
time, that indeed the arrangement simply reflected
present thinking.

By the middle of March 1946, Forrester had set up
a flow of internal information among the project engineers
that he intended would indicate what each investigator was
doing every two weeks along various of the following lines
of investigation:

Block Diagrams

Computing Circuits

Mathematic s

Mechanical, including Cockpit

4.8

Mercury Delay Lines

Storage Tube Research

5
Other Electronic Problems

He took it for granted that this beginning arrangement

would be improved upon, and it was altered as necessary

in the following years.

Of the lines of investigation indicated, only the
"Mechanical, including Cockpit" represented a contin-
uation of an earlier line of inquiry, and even that was
affected by the knowledge that devices must be developed
to convert the digital, electronic -pulse data into mechan-
ical forces and motions affecting the pilot and the cockpit.
Further, the responsive forces generated by the pilot's
movements of the mock controls must be converted back
into corresponding digital pulse data. These problems
were not impossible, but neither did established solu-
tions exist. The digital computer was too new.

Forrester's appraisal of the overall situation with
respect to computer design caused him to consider more
than one aspect of the storage problem; while mercury

4.9

delay lines as proposed for the EDVAC appeared quite
promising, so did the use of special radio tubes.
Forrester and his associates began to survey the state
of the art in this specialized area.

Computing circuits composed a category worthy of
several engineers' attention, for these circuits would
carry out the electronic operations which would perform
the appropriate arithmetic and calculating operations in
the digital mode.

It was visualized at first that the Block Diagrams
Group, the Mathematics Group, and the Electronics
Group, especially, would construct a symbiotic relation-
ship in which each would create necessary information for
the other. But in the state of engineering art as it then
existed, so unformed with respect, on the one hand, to an
Aircraft Analyzer, and, on the other hand, to a digital
computer, their relatively vast mutual ignorance imposed
contingent restraints that hobbled them together. Per-
ceiving this, the Mathematics Group sought to work its
way out of their mutual predicament of ignorance by con-
sidering ways and means of attacking the aircraft equations

4.10
that Markham and Bicknell had provided, as modified
and extended by L». Bernbaum and Bicknell. "We have
decided, " reported the head of the group, Hugh R. Boyd,
"to work on rather short specific problems and gradually
build up sufficient data and experience in numerical
methods to enable us to attack the aircraft problem
effectively. This preliminary work would also serve to

build up our knowledge of other types of problems which

7
our computer would solve effectively. "

The Electronics Group became several groups,
oriented to the things they were working with such as
circuits, pulse transformers, mercury delay lines, and
storage tubes. They were component-oriented, and they
realized that decisions from the Block Diagrams Group
would give them information about more elaborate com-
ponents and their systemic relationships. A demonstration
adder, clock pulse generator, switching arrangements, and

electrostatic tubes were among the devices under study and

• , . , . 8

construction during the spring.

The task of the Block Diagrams Group, as described

by its head and only full-time member, at the time,

Robert Everett, was "in general, to devise a complete

4.11

computer system, including definitions of all components,

interconnections of these components, [and the] sequence

9
of operations. " At the same time that the Mathematics

Group would be a source of information about computer
requirements, the Block Diagrams Group would be ascer-
taining machine computing techniques, programming
techniques, and component designs for accomplishing
computing, storing, switching, and programming.

The hindsight of experience showed the attempted
correlations of the responsibilities of these groups to
have been at once reasonable and na'ive. Had they been
mathematicians instead of engineers, the young men
involved might have placed the power and responsibility
to lead the way in the hands of the Mathematics Group.
Here, too, they would have been reasonable and naive.
But they did not, and the Block Diagrams Group became
the leader as the months wore on.

The Block Diagrams Group, meanwhile began to
analyze possible ways to proceed. By early April, with
Everett assigned to the job full time, Steve Dodd assigned
1/5 time, Pat Youtz 3/8 time, and P. Tilton 1/2 time,

4.12

the Group found open to it "a great number of system
possibilities ranging all the way from the completely
serial or sequential method described by Von Neumann,
where no two operations are performed at once, to a
completely parallel method where all operations are
carried out at once, including digit transmission. "
The latter method would be equivalent in complexity to
an analogy type solution, Everett felt. He saw that the
range of possibilities represented "a complete range of
solution time and a complete range of complexity and
duplication of equipment. Some intermediate complexity
must eventually be chosen, the criterion being that the
total equipment must be as simple as possible but still
provide the required solutions. "

Everett went on to point out that three considerations
dictated the course of action of the Block Diagrams Group.
(1) The Mathematics Group had to determine the mathe-
matical phrasing and solution procedures of the Aircraft
Analyzer equations, in order to know the "maximum
expected total of operations required in a fixed time period. "

4.13

(2) The Electronics Group had to ascertain "the time
required for a single operation. " (3) The Block Dia-
grams Group had to acquire a knowledge of components
that would "allow the most efficient paralleling of
equipment, to satisfy" the requirements the other two
Groups were working with. These strictures made it
apparent to Everett that "no final system block diagrams
can be developed for a long time, " although final designs
of components could probably be estimated closely enough
to provide these elements of the system when they were
needed.

Since explicit system parameters were as yet unavail-
able, Everett proceeded to consider the order of magnitude
of data, orders, and solution procedures a computer might
be expected to handle when coping with operations of the
Aircraft Analyzer. Acting upon the preliminary assump-
tions that he thus constructed and proceeding in the direction
indicated by the ENIAC and EDVAC enterprises of Eckert
and Mauchly, Everett envisioned a machine that would
have a total storage of about 8,200 words or less, that
would accommodate a word length of about 30 binary digits,

4.14

that would round off numbers as a fixed policy to begin
with, while the problem of errors resulting from round-
ing off would be taken up later, that would operate its
storage tubes serially, that would use a high-speed multi-
plier, that would perform input and output operations
simultaneously, and that would operate as a sequential
machine .

Everett suffered no illusions about what he and his
Group did not know; a first purpose of their early efforts
would be "to learn as much as possible about computer
techniques and problems. " At the same time, they
would be providing the Electronics Group with "prelimin-
ary specifications to enable them to better direct their

13

efforts."

Thus the young engineers under Forrester's direc-
tion spent the year of 1946 exploring possibilities, selecting
from these the arrangements, designs, requirements,
practical limits, characteristics, theoretical models, and
bench-test items they found promising. Some worked on
hardware designs. Some worked on mathematical proce-
dures that would be amenable to machine handling and

4.15

machine solution. Some worked on the problems peculiar
to creating a machine - the Analyzer and its computer -
that, to work properly, must consist of an integrated
system of component electronic and mechanical mecha-
nisms and sub-mechanisms. However efficiently and
reliably a particular circuit or subassembly might per-
form its functions when tested by itself, how would it
work when interconnected with other circuits? Would
an array of these generate sufficient "noise" - residual
currents, stray impedances and interference, back emf's,
and the like - to cause a theoretically simple arrangement
to become inordinately complex as a consequence of making
it work in practice? Especially important were the policy-
level design decisions that would give the system its basic
character. Should a storage assembly acquire its unit of
information (technically called a "word") bit by bit or
should it acquire it all at once? If mercury delay lines
were used, means of inserting information essential to
the calculating processes of the computer (interpolating)
must be provided, complicating the circuitry. Since
pulses of electric current constituted the basic signals

4.16

the computer would use, the timing and routing of these
must be finely controlled at all times. Since the digital
mode of operation meant that fresh signals were used
either to alter the character of earlier signals or to
alter the character of a patterned arrangement of
earlier signals, a "domino" effect was a prevailing fea-
ture. If just one "domino" fell the wrong way, if just
one signal were mistimed, misrouted, or were to cause
a wrong radio tube to operate, or if one tube or circuit
malfunctioned and no "back up" component were there to
compensate for the failure, then all of the rest of the
calculations to follow would be in error. In homely
analogy, the computer was an intricate array of "bucket
brigades, " and if one bucket failed to be passed on, then
the entire operation was nullified.

On the other hand, even though for want of a nail a
kingdom could be lost, reliability and coherence were
practical possibilities because of the "building -block, "
or modular, construction that was possible. A reliable
gate circuit could be inserted wherever it was needed,
like a building -block in a wall. The digital comput er

4.17

would be a more complex piece of machinery than, say,
any automobile, yet it could employ the same subme ona-
nisms over and over, as a tree does by employing not
one leaf but hundreds simultaneously. Unlike the tree,
the computer must have its modular submechanisms -
its multivibrator "flip -flops \ ] its gate circuits, etc. -
interconnected in contingent patters in such a way that
the static hook-up of tubes, wires, resistors, condensers,
diodes, and the like could accommodate and effect a
dynamic, ordered pattern of flow of radio pulses.

Long before spring, in 1946, the project engineers
had passed beyond these simple considerations, which
have been represented here in oversimplified language,
to the more sophisticated design and construction chal-
lenges of working on the detailed technical specifications.
Forrester had perceived at the start that although the
mercury-delay-line storage principle possessed many
attractive features, it might prove slow for the needs of
the Aircraft Analyzer, especially when part of a serial,
or sequential-pulse, machine. Flip-flops could be used
as storage devices, but in a simultaneous -pulse machine,

4.18

so many tubes would be involved that keeping them all
replaced and operating would be well-nigh impossible.

Electrostatic storage offered an attractive alterna-
tive principle. Various investigators in the field of
vacuum-tube research were working upon applications of
this principle, by which a minute spot on a signal plate
could be negatively or positively charged and hold, or
store, that charge long enough to be useful. The RCA
"Selectron" tube, the Williams tube (named after its
British inventor), and an electrostatic tube developed by
another MIT laboratory, the Radiation Laboratory, were
among applications that attracted Forrester's attention as
1946 wore on, and by autum he, Steve Dodd, who had been
carrying on preliminary tests, and their associates

decided to modify the MIT tube design to fit their parti -

i a 14

cular needs.

In the meantime, Everett and his group had found

compelling reasons to discard the sequential mode of

pulse operation and adopt the higher -speed, simultaneous,

or parallel, transmission of digits (pulses) among the

15
circuits of the machine. A high-speed, parallel-digit

4.19

multiplier appeared promising for the same reasons,
the most important of which was the speed of computa-
tion required if the Aircraft Simulator were to work.

Not only did the parallel- signals computer look both
promising and feasible, but also all knew that the time
was rapidly approaching when the project must "either
fish or cut bait. " Forrester had pointed out in early
June that "we are not yet in a position to decide what

must be built by next June until we know the basic prin-

17
ciples we are to use as a foundation. ..." Viewing the

performance limits within which the first computer they
proposed to build must operate, Forrester pointed out
that the contract with the Navy "calls for a model com-
puter which will, at the very minimum, demonstrate
operating principles which we plan to incorporate in the
aircraft analyzer. At most, it may become a computer
which will be useful for solving a variety of other pro-
blems. " If the latter multi-purpose type of computer
were decided upon, then, Forrester advised, it might

become necessary to extend the estimated terminal date

1 8
for Phase One of the contract beyond June 1947.

4.20

What technical information might conceivably guide
them in establishing the basic design parameters, so
that they might proceed to build their first computer?
Forrester was ready with a provisional answer: it

would depend "largely upon the information forthcoming

19
from the electrostatics field. " This policy view of

June 1946 had hardened by December into the decision
to build a pre-prototype computer, "in view of the prob-
able complexity of the prototype computer which might

20
include some 3, 000 tubes. ..." The pre-prototype, a

"simpler experimental computer, " would "test the

components of the computer and a system made up of

them, the system being capable of doing test computations. '

It would "provide a system in which to test new components

or types of operations as they become available. " It

would "check reliability and evaluate mechanical design

and maintenance problems. " And it would be operating in

six months.

When electrostatic storage tubes had been perfected,

these could be substituted for the more primitive storage

devices that initially would be provided for test purposes.

4.21

Standard electronic and relay racks accommodating
removable assembly bases (plug -in chassis) 17 inches
deep and ten inches wide, would be used. They were
readily available and permitted easy accessibility to
and testing of the hardware of which the computer would
be composed.

Project Supervisor of the pre -prototype would be
Forrester. Harris Fahnestock (who had joined the staff
earlier in 1946) would be in charge of Production.
Everett would be in charge of the Block Diagrams Divi-
sion, Leon D. Wilson would head the Computer Division,

and David R. Brown would head the Electronic Engineer -

22
ing Division. Within the already existing organization

of Project 6345, this redirection of operations was an

evolutionary phasing -in of more specialized activity; it

did not abruptly alter the entire conduct of affairs. It

represented the sharper focusing of operations that

Forrester and his associates felt was now possible after

a year of engineering research.

The extent and magnitude and detail to which their

studies had carried them have only been suggested in

4.22

this nontechnical account. Whether they had used their
time and energies wisely is difficult to determine. They
had worked hard, they had learned a great deal. They
had added carefully selected engineers to their staff, as
well as bright young graduate students looking for subjects
for Master's Theses in Electrical Engineering. When
Forrester encountered a "business-as-usual" attitude
among government suppliers of surplus equipment his
project needed, he waited until he had clear evidence the
responses were less than reasonably prompt and intell-
gent and then used it to clear up the "bottlenecks" and to
ensure that, for a while, at least, they would get better
service. This was a never-ending battle and a normal
one, with private and governmental suppliers; from the
point of view of the project workers, they were never
long in need of what they required, nor were they con-
tinually being held up by lack of funds and materials and
technical facilities.

"We got what we needed, " recalled one engineer,
"and since there was such an extensive exchange of infor-
mation going on, it was hard to get out of line or to order

4.23

something that on one else could imagine why you'd need
it. We were given our heads, but we were held account-
able. You never knew, in the early days, when Jay
[Forrester] or your supervisor would stop by to see how
you were doing. There was never any question but that
they were there to help, and there was never any ques-
tion but that they expected you to know what you were
doing. Those that didn't, somehow moved on out. Jay
was pretty good at figuring out what it was that a man
could do that would help the work along. Many of us
were going to class and had homework, and once things
really got going, we could work morning, afternoon, or
evening. You just followed the most intelligent course of
action. "

Whatever efficiency of the project is attested to by
remembered high morale, the fact remained that as the
end of 1946 approached, it began to look increasingly as

though the six -months completion date of the pre -prototype

23
computer could not possibly be met. The details simply

could not be worked out rapidly and reliably enough. In

the longer run of affairs, however, the overall progress

4.24

of the project hinged less upon any specific practical pro-
blem or accomplishment than upon the concurrent investi-
gations of many paths that appeared promising. These
investigations, from the spring of 1946 to June of 1948,
involved the exploration of both phases of Contract N5ori-
60 as outlined in Task Order I. But increasingly the
emphasis was placed upon Phase One, development and
construction of the digital computer.

In the fall of 1947, following conversations that
Forrester and Everett had been carrying on with Navy-
Special Devices technical personnel at Sands Point, the
two young engineers prepared two technical memoranda
which were studies of possible applications of the digital
computer to naval warfare. The first of these was "a
brief study of a simplified version of the anti-submarine
problem. " The second, issued two weeks later, was more
ambitious in its scope and followed naturally from the first.
It presented "in rather general terms some possibilities in
the arrangement and use of high-speed digital computers
for the analysis, evaluation and intercommunication of

4.25

24
information in an anti- submarine naval group. " To the

best of their knowledge at the time and in after years,
Forrester and Everett knew of no earlier practical engi-
neering work on how the logic of computers could be applied

25
to interpret radar data.

The two reports taken together represented an informal
proposal for practical military application of a computer the
like of which had not yet been built, although Forrester and
Everett specifically had Whirlwind in mind. "For the simpli-
fied problem selected, " they wrote in Report L-l, the
Whirlwind I computer is entirely adequate for a problem in-
volving 10 ships, 5 submarines, interconnecting radar and
sonar data, and depth charges in any number up to 20 pre-
set units and 20 proximity -fuze units in the water at one

26
time. " While the first report was interested primarily

in examining how a destroyer could acquire target data and
translate these into depth-charge firing orders by means of
a computer, the second report was concerned with the all-
important details of the sort of communication among the
ships of an anti-submarine task group that would provide
true combat information and control as the battle situation

4.26

was developing. Accordingly, the second report examined
the following example in detail: "Five surface ships and
one aircraft are illustrated with two targets, one surface,
and one submerged. All units collect such information as
they are able by the various methods noted. The computa-
tion and information system must make use of this total

27
body of information to the best possible advantage. " The

problem they then set up and explored in detail would require,

they concluded, "one -half the storage capacity and one -third

28
the operating time of WWI. "

There was no question in their minds that the computer

they were getting ready to build would be able to handle such

problems with capacity and time to spare. Both they and the

Navy Special Devices engineers enthusiastically realized

that they were contemplating a revolutionary device which

would contribute immeasurably to the efficiency and accuracy

of solving target problems in actual battle operations, but

they knew also that they could as yet only talk about "paper

operations. " Actual testing in practice lay in the problematic

future, and while they were convinced more than ever, after

4.27

these detailed studies, that they had a general -purpose
computer of a practical type truly in prospect, their
more immediate problems late in 1947 lay in the realm of
translating their ideas further into engineering designs and
their designs further into working hardware.

As their efforts came to be more and more completely
devoted to working out the engineering intricacies of the
projected computer, the Aircraft Stability and Control Anal-
yzer assumed a position of lesser importance in their minds.
It was but one example of the practical applications to which
Whirlwind might be turned, and although it still posed severe
development problems in its own right, the amount of funds
and the scale of enterprise reflected in Project Whirlwind as
well as their innovating engineering predilections produced a
"first things first" attitude that reasonably centered their
attention upon the computer itself.

Engineering development of the cockpit and its ancillary
gear for the Aircraft Stability and Control Analyzer continued
until June, 1948, when the decision was finally reached to dis-
continue that phase of the project entirely. This decision

4.28

recognized the course which the program had been follow-
ing and marked the total preoccupation of the project with
the effort to develop a general -purpose digital computer.

Upon public announcement of termination two reasons
were advanced for the decision: (1) This phase of the
total program had been carried forward as far as possible
under the existing state of the art. Further information
regarding the conversion of digital quantities to analogue
quantities was necessary; however, the research necessary
to this end could not be pursued since other phases of the
total program more urgently required the engineering and
financial resources available. (2) Continuation of Phase
One of the total program was really unnecessary, since the
pace which had been followed in the design and construction

of the simulation equipment would have resulted in its avail-

29
ability prior to completion of the computer.

The decision to discontinue the cockpit phase of the

project was not unanticipated. De -emphasis of this phase

had been accelerated throughout the winter and early spring

of 1948, and the reduction in effort had been brought to the

30
attention of both the Navy and the Institute by Forrester.

4.29

The decision was also in accord with recommendations
made by Perry Crawford in December of 1947 that the

work on the cockpit be discontinued as "not essential to

31
the program" at the time. It was only absorbing money

and engineering talent which could be applied with greater
benefit to development of the computer. Official naval
acceptance of the decision was acknowledged in August,
1948, and its necessity was justified on the grounds that
the research effort required to develop the digital com-
puter "for comprehensive real-time simulation for

32
synthetic evaluation was too enormous. " The truth of

the matter was that the Navy was running low on research

and development monies, and Special Devices personnel

33
were well aware of the fact.

The change in emphasis did not go completely un-
challenged. During the course of a conference called by
the Commander of the Office of Naval Research - which
had replaced the Office of Research and Inventions as the
parent organization for the Special Devices Division -
the question of the initial goal of the project was raised,
and some of the Navy participants expressed the hope

4.30

that the project would not "deviate too far from its
original aim of producing a high-performance facility
for analyzing proposed aircraft. " In response to these
doubts, Forrester explained that the digital computer
anticipated had never been intended to be the aircraft
analyzer, but rather a working model of the type of
computer which could be used in_the Aircraft Stability
and Control Analyzer. However, he added, it would
have limited applicability to the initial device. Through-
out his remarks, nevertheless, was the implication that
the computer he and his associates were seeking to design
and construct would be in truth a general -purpose com-
puter which in addition to scientific calculations could be

34
applied to "limited, real-time aircraft simulation. "

The decision to discontinue work on the cockpit pre-
vailed, and in October Navy Special Devices personnel

proposed that the cockpit which had been acquired earlier

35
from the Air Force be disposed of as surplus. The

following December, after the useful spare parts had

been removed, the fuselage, cockpit, and turret were

36

consigned to the scrap heap.

4.31

In the meantime, in November, 1948, the opposition
fired one more shot in defense of the aircraft analyzer.
The Mathematics Branch of the Office of Naval Research,
whose head, Dr. Mina Rees, had expressed some reser-
vations concerning Forrester's comments at the September
conference, investigated the possibility of realizing the
original purpose of the project through the use of analogue
equipment being developed under another Navy program.
The investigation, conducted by Dr. C. V. L. Smith of
the Mathematics Branch, reached a negative conclusion,
but it was a qualified negative. Dr. Smith stated in his
report that if the equations initially supplied by the Depart-
ment of Aeronautical Engineering of the Institute were to
be used, analogue equipment could not perform the compu-
tations necessary in the time required. He then proceeded
to question whether the "mathematical formulation of the
Whirlwind' problem" had not been too elaborate, thereby
opening the possibility that more simplified equations

might not only meet the requirements of the device, but

37
also permit the use of analogue computational techniques.

There the matter stood.

4.32

Subsequent events were to suggest that perhaps this
was less the final shot in defense of the analyzer than it
was the opening shot in a conflict between engineer and
mathematician that was to characterize future relations
between the MIT group and the Navy.

In retrospect it would appear that throughout this
early formative period in Project Whirlwind's history,
the Special Devices Division represented effectively both
the Institute's cause and its own as it sought and obtained
from higher Naval authority the permission and funds
necessary to change and expand the program. It is argu-
able that the Navy's acceptance of the revised program
represented a tacit, although not explicit, encouragement
of concentration of effort upon the development of a
"universal" computer rather than one peculiar to the air-
craft analyzer. If so, it would follow that the investigators
engaged in the project would feel justified in elevating the
computer research and development phase of the total
effort to primacy, subordinating the "aircraft analyzer" to
a secondary requirement to be met later if at all.

4.33

The policies developed and followed during this early
period were acceptable to Institute leadership and to Navy
leadership, and so were the improvisations and modifica-
tions of these policies. The shift of emphasis from air-
craft analyzer to universal-purpose computer was not
always destined to receive Navy endorsement, for the
times changed, the temper of the times changed, and so
did the Navy personnel. Among the factors contributing
to a deterioration of sympathetic support were reduction
in Navy research and development budgets after the war,
appearance of a new philosophy of research and develop-
ment sponsorship in the Navy, the consequent emergence
of the Office of Naval Research, and the inevitable personnel
changes in the offices designated to oversee the Navy's role
as fiscal sponsor of Project Whirlwind. These factors
caused the early rapport between Servomechanisms Labora-
tory personnel and Navy personnel to be dimmed, if not
extinguished. Unfortunately, the powerful operation of
these factors could not be checked. They increasingly
blurred and obscured the intrinsic merit and promise of
the unique Whirlwind configuration of the digital computer.

NOTES TO CHAPTER 4

1. Ltr. , J. W. Forrester to N. McL. Sage,
January 7, 1946.

2. Conference Note C-10, by J. W. Forrester,
February 11, 1946.

3. Administrative Memo A- 1, J. W. Forrester to
Engineers of Project 6345, February 27, 1946,
Subj. : "Present Status of Contractual Relations
with Navy as Regards DIC 6345. "

4. Interview, J. W. Forrester by the authors,
July 31, 1963.

5. Engineering Note E-8, Subj. : "Cockpit Program;"
Mar. 1, 1946; Administrative Memo A-3, subj.:
"Electronic Staff, " Mar. 5, 1946; Administrative
Memo A-7, subj.: "Progress Reports," Mar. 15, 1946;
Administrative Memo. A-9, subj.: "Reports, Schedules
and Meetings. "

6. Report 64, "ASCA Equations, " originally dated
October 31, 1945, revised Apr. 4, 1946 for
Project 6345.

7. Engineering Note E-14, subj.: "Mathematics Group,"
Apr. 12, 1946.

8. Administrative Memo A-3, subj. : "Electronic Staff, "
Mar. 5, 1946. Conference Note C-12, subj.: "General
Meeting of Staff Members of Project 6345, " June 10, 1946.

9. Engineering Note E-13, subj.: "Block Diagrams Group,"
p. 1, Apr. 3, 1946.

10. Ibid. , p. 1.

1 1. Ibid. , p. 2.

12. Ibid. , pp. 3-4.

NOTES TO CHAPTER 4 (CONTINUED)

13. Ibid. , pp. 4-5.

14. Conference Note C-19, subj. : "Parti - Report of
Lab. Work - S. Dodd;" "Part II, Report of Proposed
Storage Tube Program - J. W. Forrester, "
October 23, 1946.

15. Conference Note C-22, subj. : "Discussion of a
Parallel Computer, " November 6, 1946.

16. Conference Note C-24, subj. : "A High-Speed
Parallel Digit Multiplier, " Nov. 20, 1946.

17. Conference Note C-12, subj. : "General Meeting of
Staff Members of Project 6345, " p. 2, June 10, 1946.

18. Ibid.

19. Ibid.

20. Conference Note C-25, subj. : "Pre- Prototype
Computer, " Dec. 2, 1946.

21. Ibid.

22. Ibid. Memorandum M- 43, subj.: "Pre-prototype
Status, Report," Dec. 10, 1046; memoM-52, subj.:
"Notes and Block Diagrams for the Pre- Prototype
Compxiter, " Dec. 27, 1946.

23. Memo M-47, Subj. : "Pre-prototype Computer
Meeting, " Dec. 17, 1946.

24. These were the first of the L-Series of reports that
the Project began to issue. Report L-l appeared as
Memorandum M-108 but was shortly changed to:
Report L-l, J. W. Forrester and R. R. Everett to
Director, Special Devices Center, subj. : "Digital
Computation for Anti-submarine Problem, "
October 1, 1947. Report L-2 appeared as:
Limited Distribution Memorandum L-2, J. W.

NOTES TO CHAPTER 4 (CONTINUED)

Forrester and R. R. Everett to Director, Special
Devices Center, subj.: "Information System of
Interconnected Digital Computers, " Oct. 15, 1947.
Although entitled "Memorandum, " the items in this
series rapidly became known as "L-Reports. "
The "Special Devices Center, " referred to above,
replaced the "Special Devices Division" in the
reorganization accompanying the creation of the
Office of Naval Research; as SDD had reported
to the Office of Research and Inventions, so did
SDC report to ONR.

25. Interview, J. W. Forrester and R. R. Everett by
the authors, October 26, 1967.

26. Report L-l, p. 1.

27. Report L-2, p. 2.

28. Report L-2, p. 12.

29. Project Whirlwind Summary Report No. 9. , June 1948,
pp. 12-3.

30. Ltr., J. W. Forrester to N. Sage, subj. : "Amend-
ment No. 4 to Project Whirlwind Contract N5ori 60, "
Feb. 2, 1948; Ltr., J. W. Forrester to Director,
SDD, att'n Charles Doersam, Mar. 2, 1948.

31. Memorandum, Perry Crawford, Jr. , to Director,
SDC, Subj.: "Whirlwind Program, " Dec. 18, 1947.

32. Proposed Work Description for Whirlwind Pro-
curement, August 16, 1948, corrected by Perry
Crawford, August 16, 1948.

33. Interview, C. R. Wieser by the authors,
June 16, 196 5.

NOTES TO CHAPTER 4 (CONTINUED)

34. Perry Crawford, Jr. , Memorandum for the files,
subj. : "Conference on Project Whirlwind Held at
Navy Department, 22 September 1948," Nov. 2, 1948.

35. Information on the acquisition of the Cockpit is
contained in: Ltr. , Noel Gayler, Cmdr, USN, to
CG. AMC, Wright Field, Sept. 18, 1947.

36. Memorandum, C. H. Doersam, Jr. , Computer
Section to Director, SDC, subj. : "B-24 Fuselage

for Project Whirlwind; Disposition of," Oct. 25, 1948;
memorandum, Survey and Surplus Property Review
Board to Head, Building and Ground Units, subj. :
"Surplus Property, Instructions for Disposal of, "
Dec. 3, 1948.

37. Memorandum, Code 424 (Fred D. Rigby) to Code 100,
subj. : "Report from C. V. L. Smith to Head of
Mathematics Branch, subj. : "Recommendations
Concerning the Realization of the . . . , " Nov. 18, 1948.

Chapter Five
PRESSURE FROM ONR

It was equally easy to take the view at the start of
1947 that Project Whirlwind was making due progress or
that it was falling behind, depending upon the expectations
of the observer. In either case, the selection of a proper
scale against which to measure the activities of the project
remained a complicated, intuitive, highly subjective, and
obscure task of judgment, further complicated by the common,
joint practice of establishing goals and schedules to be met
that the Massachusetts Institute of Technology and the Navy
followed. In this respect, Project Whirlwind was like many,
if not most, research and development projects. Goals and
schedules had been set, providing a time-table for exploring
the unknown and the partly known; since the time-table was
the product of mixed ignorance and knowledge, it is not
surprising that subsequent investigation showed these goals
to be less attainable and more remote than earlier had been
thought. Revised goals were necessary, and these called
for further investigation, further research and development

5. 1

5.2

effort that yielded additional information. Inevitably, some
of the new information, in its turn, further modified, trans-
formed, or even destroyed some of the revised goals.

Hindsight in later years might tell whether there had
been progress and of what kind, but here, too, the proper
scale of measurement is not easy to select. On the one
hand, the engineers in the project, their counterpart Navy
Special Devices program managers, administrative superiors
at the Institute and in the Navy's Office of Naval Research,
which had replaced the Office of Research and Inventions
during the latter half of 1946, all had the opportunity to
contemplate the impressive record of past achievements
of MIT, of its Division of Industrial Cooperation under Nat
Sage and of its Servomechanisms Laboratory under Gordon
Brown. On the other hand, they could survey apprehensively
the still-unsolved problems and the new, relatively formless
state of the art with which the MIT engineers were struggling
in the digital computer realm.

5. 3

Forecasts which engineers and administrators
cautiously generate while they are contemplating trouble-
some problems rarely agree with those which they
optimistically foresee while they are reviewing past
accomplishments, and any analysis that attempts to
combine the virtues of both runs the risk of appearing
to present confusing, vague, and contradictory statements
at best and outright doubletalk at worst. The evidence
that is most convincing to the insider is least convincing
to the outsider. The conclusions resting upon such
internal evidence are most persuasive to those familiar
with the evidence and least persuasive to those who,
viewing it from without, lack the feel of its pulse and the
sense of its past and present color that provide good
rate-of-progress information.

It was from this sort of predicament that Jay
Forrester sought to extricate himself when writing a
semi-annual review of the status of the Whirlwind
contract as of January 1947. While he and his associates
were able to keep their heads above water, the current

5.4

of events continued inexorably, slowly, steadily to carry
the Project toward certain shoals and reefs that were
forming as a consequence of actions taken by the Navy to
reorganize its practices and policies for supervising and
funding research and development projects.

Mindful of the two Tasks that had been written into
Contract N5ori-60 a year earlier, Forrester was willing
to admit to Special Devices program managers and their
superiors in the new Office of Naval Research that Phase
Two, the construction of a prototype computer, would
commence not in July 1947, as planned, but in January 1948,
in order to allow sufficient development- time for the pre-

prototype and thus establish, with sufficient firmness to

2
proceed, the configuration of the Phase Two device.

The computation speeds of the machine would have

to be "well above those originally anticipated, " and although

the general nature of the computer block diagrams had been

established under Everett's leadership, much new work

involving "the advancement of electronic techniques in the

fields of video circuits, electronic switching, trigger

circuits, and pulse transformers" lay ahead. Further-

5. 5

more, since reliability of operation in a computer hooked
up to an aircraft simulator was crucial, "checking and
trouble- shooting circuits similar to those required in the

final electronic computer" should be incorporated, and

3
time should be allowed for this. The Institute felt, said

Forrester -- and by this he meant himself and his associates

working on Project DIC-6345 in the Servomechanisms

Laboratory and implied also Professor Gordon Brown,

his superior, and Nat Sage, his administrative supporter

and protector, who received a copy of the letter, -- the

Institute felt that the pre -prototype ought to possess the

operating speeds and approximate circuits the final machine

would feature.

Forrester believed that in view of existing circumstances,

the design of the pre-prototype could be firmly established by

the end of the next eight or nine months, and construction

of its many parts could be completed three or four months

later (having begun well before October). Such would be

the state of the Project at the end of 1947, and it would

permit the pre-prototype to be assembled and put into

5.6

4
preliminary operation "early in 1948. "

He was careful to exclude the electrostatic storage
tubes from this schedule. They might be available in
time, but sufficient data were "not now available to make
firm time estimates, " so manual-switch storage and
flip-flop storage would be incorporated until such time
as the tubes became available. About 25, 000 binary
digits of future electrostatic storage were called for in
the plans, but they were still only in the plans.

To accomplish the revised, stretched- out schedule,
a level of expenditure of $30, 000 per month would be
required for the coming year, and most of these expenses
would be charged to the pre-prototype of Phase One of
the contract. As both MIT and SDC representatives well
knew, Phase One had never been intended to "define the
nature or extent of this pre-prototype computer. "
Forrester could nevertheless assure the Navy that "the
project is now prepared to embark upon the specific

system design of a pre-prototype electronic computer

5
which is the end objective of Phase 1. "

5.7

Benefitting from the past year's researches, the
pre-prototype would employ parallel, or simultaneous,
transmission of digits. Block diagrams that Everett
had developed for a serial- transmission computer --
inspired originally by the proposed EDVAC machine --
had convinced the engineers that, despite relatively
simple and easy-to-maintain circuits, such a device
would be too slow. So block diagrams for a faster,
parallel computer had been developed. Registers large
enough to accommodate 16 binary digits would be employed.
"Sixteen digits are considered sufficient for testing and
demonstrating electronic operation and for a certain few
investigations into the mathematical applications of digital
computers, " Forrester observed, but realizing how much
more useful in mathematical investigation such an instrument
might be if it could handle larger numbers, he added that the
computer would be so designed as to carry out its operations
"in multiples of 16 digits in length, " specifically, 32 binary
digits.

5.8

It was a well- composed letter, and it said more than
is indicated here. It was packed with information, presenting
the good news with the less than satisfactory and putting the
latter softly, so as not to disturb. Nevertheless, it was a
letter that said progress was slower than had been scheduled,
and when one paused and reflected upon just what sort of
slow progress it was, one could see that it was progress to
the tune of $305,000 already spent, progress to the further
tune of another $200, 000 anticipated for the next six months,
and progress to still another tune of $528, 000 for the year
after that -- over a million dollars -- and no assurance
when the storage tubes would be ready. It promised ultra-
high computer speeds, only 16-digit operation to begin with,
and some kind of storage some time. For a million dollars
and more !

This state of affairs represented not unreasonable
progress to any Navy program manager who had been
involved at the start of the engineering project to develop
an Aircraft Stability and Control Analyzer, nor was Lt

5.9

cause for special concern to those, like deFlorez, Gratiot,
or Crawford, who had entertained a rather visionary and
aggressively dynamic engineering philosophy with regard to
certain technical developments they considered desirable
and feasible. But just as Project Whirlwind's behavior could
be accounted for by the very character of the research and
development process it was engaged in, so could the Navy's
changing attitude be explained by fundamental organizational
and policy changes which were taking place within the Navy.
These were generated by circumstances that historically
and genetically had nothing to do with aircraft analyzers
and digital computers, and they were too profound to be
affected by the small influence that the Navy Special Devices
personnel and MIT's Project Whirlwind engineers could
exert. The accumulating impact of these changes the
Special Devices Center and Project Whirlwind, put in
strongest terms, emasculated the former and drove the
latter to the wall. Put in milder terms, these changes
inevitably effected a major reassessment of some of the
projects in which the Navy and civilian advanced research
and development teams were jointly engaged -- and one of
these projects happened to be Project Whirlwind.

5. 10

At the time of the inception of the Aircraft Analyzer
program at MIT in 1944, Navy supervision and funding of
the program had been a primary responsibility of the Special
Devices Division, initially organized as a branch of the
Bureau of Aeronautics, but transferred along with other
Navy research and development facilities and organizations
to the Office of Research and Inventions in May of 1945,
and subsequently, in August of 1946, to the newly created
Office of Naval Research. Under both the Bureau of
Aeronautics and the Office of Research and Inventions, the
Division had been permitted a wide latitude of authority
and freedom of action. Once under the control of the Office
of Naval Research, however, the Division was phased out
and its facility at Sands Point, Port Washington, New York
was designated as the Special Devices Center of the Office
of Naval Research. Until February of 1949, nevertheless,
Navy responsibility for the supervision of Project Whirlwind

was to remain with the Special Devices Center, thus assuring

7
a continuity of supervision up to that time.

5. 11

Prior to the appearance of the Office of Naval
Research there had developed between the engineers of
Project Whirlwind and the engineers of the Special Devices
Division a reciprocal confidence and sympathy which was
to decrease proportionately to the increase in ONR's
exercise of authority over the Center and its programs.
There had been occasional areas of disagreement between
the two groups, but relations had been basically sympathetic
and understanding, the product of a rapport grounded in
the engineering orientation of the two groups and in a
common parenthood of the Aircraft Stability and Control
Analyzer. The relatively harmonious relations which
had been established early in the program were given
additional strength and substance when Perry Crawford, Jr.
left the Massachusetts Institute of Technology in October
of 1945 to join the staff of the Special Devices Division.
Crawford, who subsequently suggested the use of the
digital computer as a solution to the real-time problem
which was besetting Forrester and his colleagues,
brought to SDD additional familiarity with the Project,
but more importantly, he brought with him an imagina-

5. 12

tive and enthusiastic confidence in the potential utility
and versatility of the digital computer. As head of the
Special Devices Center's computer section, Crawford
was to prove an imaginative, able, and influential ally

o

to Project Whirlwind until ONR took full command.

As SDC became more and more the instrument of
ONR, however, the relations between Project Whirlwind
and the Center became increasingly strained and critical,
to such a degree that in the winter of 1947, Forrester
even questioned the Center's competence "to provide
the proper administrative, technical and financial
assistance to the work and to properly relate the interests
of all Navy groups. " Undoubtedly, Forrester was
becoming increasingly restive under the more critical
supervision which was emanating from SDC in response
to the increasing pressures generated by ONR. The

earlier rapport was being submerged by the tensions

o
which were created as ONR asserted its authority.

5. 13

The years 1947 through 1949 were difficult years
for Forrester and Project Whirlwind, for in addition
to the increasing tempo and severity of Navy criticism,
Project Whirlwind found itself under closer and more
penetrating scrutiny by the Institute's top administration.
Project Whirlwind had become a source of contention,
caught up in the struggle which accompanied ONR's
efforts to implement the authority inherent within its
enabling legislation. It was caught up also in the struggle
between mathematician and engineer which accompanied
the pioneering research and development phase of digital
computation. Finally, it was caught up in the struggle
over funds which accompanied post-war retrenchment.

Jay Forrester's direction of the Project also
became involved in the controversy. Dedicated to
Project Whirlwind and determined to secure its success,
Forrester aggressively and single-mindedly pursued the
course which he believed would most quickly reach that
end. Without doubt, his aggressiveness and determina-
tion offended many, but without this sense of purpose

5. 14

behind it, the Project could very likely have failed.
His superiors both within MIT and the Navy no doubt
were pleased by his determination to do the best job
possible on what he considered to be "one of the most
important development jobs in the country, " for he was
convinced that computers promised rewards to the
military as great if not greater than radar. But his
attitude toward costs could not fail to be disturbing
because those which others regarded as expenses,
imposing an upper limit, he seemed to consider as
productive investments, as means to an end, rather
than determinants of level of effort. His apparently
cavalier attitude toward costs was doubly disturbing
because of his youth and because he apparently failed
to communicate effectively his rationale or philosophy,
if indeed he had one in that regard, to cost-conscious
Navy supervisors compelled to stay within limited,
peacetime budgets. They were dismayed, not reassured,
by his conviction "that the facilities and funds needed
to do a job are subordinate to getting the job done as

5. 15

quickly as technical progress permits."

The rate of progress Forrester claimed for the
program did not go unchallenged. Thus, his letter of
January 1947, carefully composed though it was, as
we have just seen, nevertheless left the program
vulnerable. For it compared progress accomplished
to goals set and permitted critical eyes to find the
progress wanting. It suggested the goals-- the time
schedule --be modified, but this suggestion raised
again the impression that progress had not been
satisfactory. And indeed, judged by the goals set
earl ier, the progress had not been satisfactory.
So Navy programmers could ask, was this indeed the
case? Had the goals been unrealistic? Or were the
capacities of the researchers inadequate? First they
were going to build a simulator. Now all that was
discussed was a computer, and this wasn't even the
computer, or the sort of computer, that the project
had set out to build. Were they eager young men who
had gotten beyond their depth and didn't realize it --

5. 16

wouldn't realize it -- yet? Two and a half weeks after
Forrester sent in his letter, the head of the Naval Research
Advisory Committee, a civilian scientist, spent an hour
and a half on an inspection visit.

Dr. Mina S. Rees, Head of the Mathematics Branch
of ONR, was of the opinion that the "consensus of visitors
to the project is that there is too much talk and not enough
machine. " To the mathematician who visited the Project
and who lacked understanding of the engineering problems
involved, this comment seemed only too self-evident and
accurate. Also, criticisms voiced of Forrester and his
project could on occasion be extremely harsh and extreme,
reflecting as one observer noted, "the personal animosity
which is widespread in the computer development field
and especially as regards Mr. Forrester. " It is not
impossible that such criticisms, even when discounted
for their exaggeration, were influential even though not
sufficient by themselves in shaping Mina Rees' view of
Project Whirlwind and its staff.

5. 17

Without doubt, Forrester and his associates were
operating in a very competitive field and one in which
the mathematician was a powerful if not a dominant
influence. Young, inexperienced, and unknown engineers,
they were matching skills and abilities with men of known
stature and status such as John von Neumann of the
Institute for Advanced Studies, Howard Aiken of Harvard,
J. P. Eckert and J. W. Mauchly of the University of
Pennsylvania, G. R. Stibitz and S. B. Williams of the
Bell Telephone Laboratories, and M. V. Wilkes of
Cambridge, England, just to mention a few.

The electronic automatic sequence control machine
was in its early conceptual stage. These young engineers
were seeking not to refine an already existing device,
but rather to design, develop, and construct an entirely
new one. In short, they were converting a concept into
an electrical system embodied in a piece of tangible
hardware. If the Whirlwind engineers had not been
operating within the protective womb of MIT, it is
altogether conceivable that the Project would have been
terminated by the Navy, particularly after ONR had

5. 18

assumed primary responsibility for Navy research and
development. The mathematicians of ONR, enamored of
the computer as a scientific instrument of rapid calculation,
failed to recognize its potential as a command and control
center as early advocated by Forrester and Crawford.
The engineers of Project Whirlwind and SDC, concerned
primarily with application to military needs rather than
development of theoretical concepts, saw it as an
instrument primarily adapted to facilitate human control

of events in the physical world and only secondarily

12
intended as a mathematician's tool.

The misgivings expressed by Mina Rees were not

hers alone, for they were shared by her colleagues within

the Mathematics Branch. The mathematicians were

concerned because they believed that neither Forrester

nor any of his associates actively engaged on the Project

possessed the "mathematical competence needed in the

13
design of a new type digital computer. " Such misgivings

were not a sudden development, nor were they allayed by

Forrester's semi-annual review submitted at the end of

5. 19

January, 1947. They led in February to a visit by
Warren Weaver, then Head of the Naval Research
Advisory Committee, to the Servomechanisms Laboratory
at MIT to investigate the Project. Later Weaver was
also to visit SDC at Sands Point. After visiting the
Institute, Weaver in his comments to Mina Rees
expressed no major criticisms, or praise either, of
the Project or the personnel engaged upon it, but he
did raise some very penetrating questions without
providing the answers. Included among the questions
which were concerned primarily with the nature and
purpose of the Project was one pertaining to the quality
of the mathematics in the program. Weaver wondered
if it matched their "excellent physics and engineering? "
Subsequently, after visiting SDC and writing with greater
retrospection, Weaver observed that neither achievement
nor progress could be measured by a single visit. His
conversation with Forrester had left him, he observed,
with the belief that there was some confusion whether
Whirlwind was really "a simulator or a general-purpose

5.20

computer, " a belief caused by Forrester's description
of Whirlwind at one point in the conversation as a
general-purpose computer. But when pressed by Weaver
to explain how it would handle certain scientific calcula-
tions, Forrester evaded a direct answer by describing
it as a "fire-control" computer. As Weaver put it, was
the Project failing to be good biscuits by trying to be
cake? Crucial here were the value judgments to be
applied; it was easy indeed to make invidious comparisons
of engineering simulators and biscuits, on the one hand,
with scientific mathematical machines and cake, on the
other, and Weaver was wary of rendering such judgments
even while phrasing the problem in perhaps suggestive
terms. The strongest impression he gained from his
visits to MIT and to SDC was that both Forrester and
Crawford were extremely competent and able, and that

the Whirlwind staff was "well organized, enthusiastic

14
and hard at work. "

5.21

Weaver's visit to the Servomechanism Laboratory
at MIT and his subsequent visit to SDC may have been
purely coincidental or part of a general study he was
making of the Navy's research and development program,
but coming on the heels of each other, they strongly
suggest that he was investigating the Project within its
total context, seeking to determine not only the implementa-
tion of the program at MIT, but also its direction by SDC.
Mina Rees and her colleagues were concerned about SDC,
its relations with ONR, and the guidance and direction it
was providing Project Whirlwind. This concern, the
Computer Section of SDC, understandably, felt led to an

improper interference in its area of authority, but the

15
balance of power was shifting to the Mathematics Branch.

During the course of a discussion over the establishment
at Sands Point of a "simulation facility, " using Whirlwind II
(the projected second-generation computer) as its informa-
tion and control center, Mina Rees while expressing
approval had some reservations lest she was "relinquishing
some responsibilities that properly belong to the

5.22

Mathematics Section. " In addition, in an aside to Perry-
Crawford, she questioned if the Center were not engaging
in "empire building. "

If Mina Rees and her colleagues had hoped through
Weaver's visits to obtain evidence which would support
their efforts to curb the Project --or even destroy it,
as some of the junior members of Project Whirlwind
charged in retrospect -- they were disappointed. On
the other hand, his comments did not still their
apprehensions.

It is doubtful, however, if Mina Rees and her
associates sought to destroy the Project. Certainly,
they sought to bring it under firm control, to orient it
properly, for they were seriously concerned about the
program which, they believed, had merit but lacked
direction and purpose. In the fall of 1957, Mina Rees
believed that the Project possessed real and tangible
possibilities, particularly for "scientific" computation,
and even if it failed to attain complete success, "a

5.23

substantial contribution to the art" would have been made
and "the money invested . . . worthwhile. " The money was
also, one might add, considerable in amount, a fact which
seriously disturbed ONR, as events were to prove. Between
the inception of the program and the assumption of control
by the Mathematics Branch of ONR, the estimated costs
had more than doubled and threatened to continue to mount,
and the schedule had slipped by some twelve months, yet

the original purpose of the program contractually remained

4-u 17

the same.

The pot continued to simmer, even if it did not boil,

the discontent of the mathematicians providing a steady

source of heat. They continued disturbed by the Project's

lack of that which they regarded as competent mathematical

talent essential to a well-ordered, properly organized

computer program. For them the ideal electronic digital

computer program was the one at the Institute for Advanced

Studies under the direction of Dr. John von Neumann.

Persistently, they compared the two programs, asking how

5.24

Whirlwind differed from the IAS computer. If the two
devices did not differ significantly, then why was
Whirlwind costing so much more? Persistently, also,
they asked why Whirlwind was being designed and built
as a general-purpose computer if its primary application
was to be simulation. These specific questions were
raised at an ONR conference in October 1947, accompained
by the charges that Project Whirlwind lacked essential
mathematical competence, that no effective analysis of
the functions of Whirlwind had been prepared, that the
status of the storage tube program had been exaggerated,
and that even within the MIT community, the Project
was under fire for lack of interdepartmental cooperation
and for its unsatisfactory progress rate.

Responding to these specific questions and charges,
which obviously contained the implication that SDC had
been remiss in its direction of Project Whirlwind,
Crawford recommended that Professor Francis J. Murray
of Columbia University be retained to evaluate the

5.25

"mathematical competence indicated by the work to date"
and to make a comparison between Whirlwind I and the
computer von Neumann was developing at Princeton.
In addition, Project Whirlwind 1 s directors should prepare
"detailed information concerning the components designed
for Whirlwind I and the design of the Whirlwind I system. "
Until the information requested was furnished and the
decision was reached that the program was indeed valuable,
he recommended that no further consideration be given to
the financing of Whirlwind II. Crawford's last recommenda-
tion may have accurately reflected his own annoyance and
misgivings, but certainly it mirrored the opinion of some
within the upper echelons of ONR, and implied that the

Office was threatening the use of its ultimate weapon --

19
the power of the purse --to bring the Project into line.

In order to dampen the heat persistently emanating

from ONR, Forrester followed two courses. To meet

the chronic objections, he prepared with his staff, upoii

the recommendation of Captain George M. O'Rear of

SDC, a twenty- two volume administrative and technical

5.26

summary of Project Whirlwind since its inception in 1944,

setting forth in detail the changes made in the purpose and

20
nature of the program and the reasons for them. This

report, he hoped, would explain away ONR's objections

and serve as a compendium to provide answers to any

future questions the Project's critics might ask. The

questions and charges which had been made at the October

conference at ONR were answered separately and in

specific detail.

Comparing Whirlwind I to the von Neumann computer,

Forrester argued the former was faster, more applicable

to Navy needs, and further advanced in design and

construction. Comparative costs could not be determined,

he noted, since von Neumann had no cost estimates for

his finished device; however, because of "final design

refinements and the more finished packaging, " Whirlwind's

final costs would probably exceed those of the IAS computer

by a margin greater than the two-to-one ratio forecast

by ONR. His critics, he suggested, evidenced a real

5.27

lack of understanding of "the simulation and control field
and . . . the meaning of a general purpose computer" when
they sought to make Whirlwind one or the other, for the
complexities of simulation demanded a flexibility which
permitted a wide variety of uses. The storage tube
development program was difficult and complex, but
one which had always been frankly and candidly discussed
without exaggeration.

In denial of the charge that interdepartmental
cooperation was lacking, Forrester cited instances in
which other departments had cooperated by making either
personnel or facilities available. Within his own department,
Electrical Engineering, a separate research program --
supported by the Rockefeller Foundation --in digital
computation had been discontinued to permit consolidation
of the two staffs in order to make the total effort more
effective. All in all, Forrester argued, because of the
immense importance of electronic digital computation,
M.IT had rendered more aid to the Project than the Navy

5.28

had a reasonable right to expect, and furthermore, this
assistance had been given despite heavy teaching and
research commitments.

Despite Forrester's disclaimers, supported as
they were by cited cases, there was a continuing feeling
that cooperation, if not lacking, was limited. Forrester,
belatedly perhaps, had requested assistance from other

departments, but the indications were that they had not

22

responded enthusiastically. Beyond the usual obstacles --

other commitments, lack of interest, etc. -- one significant
impediment to cooperation was without doubt the classified

nature of the Project, a barrier which Forrester found to

23
be a continuing problem. Professor Samuel Caldwell,

who had been working on the research program supported

by the Rockefeller Foundation, refused to work with

Project Whirlwind so long as it was subject to military

security restrictions. He would work only "on research

concerning electronic computing that will freely serve

all of science, " a view which was shared by many of his

colleagues.

5.29

Nevertheless, Forrester did have a measure of
assistance and cooperation from the Mathematics
Department. Professor Philip Franklin of that
department was dividing his time between departmental
and Project Whirlwind duties at the time that Crawford
called for an inspection visit by Professor Murray of
Columbia. Together with two full-time members of the
Project, Franklin constituted its Mathematics Section.
The effort put in by this group and by others working on
mathematical problems In the Project "would represent
a larger staff than available for the entire engineering
activity of the Institute for Advance Study computer" if
it included all who performed mathematical functions
within the program, Forrester pointed out. Both he
and his critics knew that this was an organizational
procedure, as well as a legitimate way of interpreting

program operations, that was not restricted to the

25
Whirlwind Project.

5.30

Although the Project was not emphasizing mathe-
matics as much as ONR felt was necessary, it was

pursuing research in pure and applied mathematics

26
related to the computer. At the same time, mathe-
matics that was not directly pertinent to the engineering
development of the hybrid, practical, general purpose,
science and engineering instrument that Forrester and
Everett visualized tended to be subordinated. Forrester
felt sufficiently vulnerable to ONR's criticisms to be
goaded into further defensive action following Murray's
visit, which occurred on November 8; four days later,
Project Whirlwind coincidentally published a memorandum
by Franklin surveying in some seven pages of single-
spaced typescript the Project's mathematical program,

27

both accomplished and planned. In addition, within

the month Forrester was planning to enlarge both the
mathematics staff and program, subject to Navy approval
indicated by adjustment of the contract "to cover continuing
basic research programs.

5. 31

The Project's activities were by no means confined
to responding to the external pressures generated by ONR's
persisting and sceptical scruting. Indeed, Forrester
shielded his engineers, so far as he was able, from the
outside alarms so that they might continue their research
activities with as little interruption as possible. In the

fall of 1946 they had begun looking actively for building

29
space to house the project pre-protype computer.

By March the firm of Jackson and Moreland, Engineers,

headed by Edward L. Moreland, Frank M. Carbert,

and Ralph D. Booth, had estimated that the accommodations

specified would cost about $770, 000 if incorporated, as

proposed, into the projected Navy Supersonic Wind Tunnel

Laboratory on the campus by extending the office section

of the Laboratory "three additional floors, making this

building a four- story building in order to house the

30
Servo-Mechanism Laboratory. "

Forrester allowed himself a growth factor in a

report to SDC in April on the matter. The Supersonic

Laboratory accommodations requested would include

5. 32

enough space, he felt, "for development and operation

31
of the final Whirlwind computer. " He could not yet

make a report, he said, on the alternative of "purchase
or rental of an existing building. "

The growth factor assumed more explicit form
in a letter to Perry Crawford near the end of April,
in which Forrester confirmed earlier verbal discussions,
for the record, in a way typical of the degree of coopera-
tion that had become characteristic of relations with SDC
and that soon was to disappear as the Mathematics Branch
of ONR assumed greater authority. Phase 1 would be
extended to June 1948 because "a reevaluation of progress
and time schedules" indicated that more research and

development time would be needed "prior to design of

32
Whirlwind 1." This was not a "stretch-out" represent-
ing reduced effort, however, because "the scope of the
Whirlwind I computer is considerably more extensive
than originally planned and will require an additional
six months' time. Since the computer will be more

5. 33

nearly like Whirlwind II than originally anticipated the
design of Whirlwind I will appreciably ease the design
and construction problem of Whirlwind II. " Looking
ahead, Forrester drew attention to developments that
both MIT and SDC viewed at that time as reasonable
projections: "It is anticipated that Task II involving
the construction of Whirlwind II will overlap somewhat
the end of Phase 2 of Task Order 1 covering the system
design and that steps will be taken as soon as possible
to formulate Task II. " A different future was in store,
however. Whirlwind II was never built, and in its place
appeared a more elaborate machine than anyone was
then planning on, the ANFSG-7.

As a matter of policy, Forrester deliberately
stopped referring to the "pre-prototype" in external
correspondence that spring; as a matter of custom,
"pre-prototype" yielded to "Whirlwind" in the
Laboratory as the summer wore on. In the meantime,
further investigation by Jackson and Moreland

5. 34

revealed that the earlier estimate of building costs had
been too low, and the Supersonic Laboratory became
less attractive as time and cost schedules indicated
an already-existing building might be more feasible.
Before the end of August the Barta Building, located on
Massachusetts Avenue close by the MIT Campus came

under serious consideration, and it was the Barta
Building that became the home of the Whirlwind computer.

The technical appraisals undertaken by Forrester,
Everett, Fahne stock, Boyd and other engineers in the
Project during 1946 had indicated that problems of
engineering reduction-to-practice were least trouble-
some in the areas of information input and output and
most troublesome in the area of storage. The Project
leaders became convinced early that fast internal
storage organized in easy-to-add-onto units was

essential, and they devoted their efforts particularly

33
to electrostatic storage. Input and output problems

they were willing to let the Navy Special Devices Center

5. 35

contract for separately, and by autumn of 1946 Eastman
Kodak was involved in providing "equipment for the
preparation of input films from a manually operated
key board as well as output recording devices and

mechanisms for reinserting output data into the input

34
of the computer. " As matters turned out, the Eastman

equipment, using minute clear or opaque spots on 35 mm.
film to represent binary digits to be implanted or read by
cathode ray tubes and associated photosensitive tubes,
was never perfected for Project Whirlwind, and other
input- output techniques brought forward by the industry-
wide advancing state of the computer art were employed
instead.

During 1947 the quota of graduate students employed
as research assistants rose from eight to twelve and

then to fifteen. By the end of October, Forrester was

35
asking for twenty for the next year. Most of these

were working towards their Master's Degree and carried

5. 36

out or assisted in special investigations that added to the
Laboratory's pooled knowledge in a modest and detailed
way while providing the subject for a Master's thesis or
occasionally a doctoral dissertation. The practice of
biinging students into the Laboratory continued as long
as it remained on the campus and geographically separate
from the MIT subsidiary it later joined, the Lincoln
Laboratory located in nearby Bedford. Not only were
Forrester and his assistants continuing Gordon Brown's
policy with regard to students, but also they found the
campus relationship invaluable in providing a small but
growing pool of first-class engineering talent which in
later years was to spread out into the growing computer
industry. While these students gave their best efforts
to the Project, often continuing on the staff after obtaining
their degrees, the Project in return gave them the
experience that put many of them a professional jump
ahead of their contemporaries.

5. 37

While technical work proceeded apace, as Warren
Weaver, Professor Murray, and other visitors observed,
the Project leaders presided over the expanding activity,
moving from details to overviews to analysis of how the
work was proceeding on many fronts, and back to details.
While Everett, for example, spent more of his time on
the complex problems of logical circuitry and attended
to the details of creating and maintaining an integrated
system of working components as research phased into
advanced design and design into projected hardware,
Forrester occupied himself with internal and external
organizational details and with building and maintaining
a high- spirited, hard-working organization. Supported
by Nat Sage's office, he selected a subcontractor to
fabricate the hardware, the racks, the panels -- the
form and substance itself -- of Whirlwind I, Sylvania
Electric Products Company of Boston took the contract
with MIT during the latter half of 1947 and went to work
building the items to the requirements and specifications
of the Whirlwind staff.

5. 3i

The Mathematics Branch of ONR endorsed these
developments even while preserving its apprehensions
over the basic direction and purpose of the project.
That there was much activity and increasing amounts
of money being spent at Project Whirlwind was no
guarantee, after all, that the money was being wisely
or well spent. What was really going on in the Servo-
mechanisms Laboratory? The Mathematics Branch
could never share SDC's confidence. The twenty-two
volumes of Summary Report Number Two, for all
their impressive and informative detail, were but
another manifestation of the peculiar style in which
Forrester's operation proceeded to go its own way,
have its own way, and -- for all Mina Rees, C. V. L.
Smith, and their associates could tell --be heading
for a spectacular fall in its own way.

ONR, consequently, welcomed Crawford's
suggestion that Francis J. Murray of Columbia
University be asked to look into the situation and
deliver a report free of the modulated yet enthusiastic

5.39

bias to be expected in the Project's own Summary Reports.

Murray, as has been remarked, visited the Project
on November 8, 1947. He was an associate professor of
mathematics who possessed both classroom and laboratory
experience in computers. He had agreed to undertake

the task Crawford proposed in his memorandum discussing
the ONR conference of October, and accompained by
representatives of SDC, including Perry Crawford, he
conferred with Forrester, Everett, and Philip Franklin
of the MIT Mathematics Department. Both professional
courtesy and official responsibility required Professor
Franklin's attendance at the conference. His presence
also served to counter the ONR charge of inadequate
attention to mathematics, a consideration Forrester was
not likely to overlook. A week later the Cambridge
conferees, minus Franklin, travelled to Princeton to
meet with Professors John von Neumann and H. H.
Goldstine for a discussion of the IAS computer program.

Within the following week, Murray had finished his

37
report and submitted it to the Director, SDC.

5.40

In his report, Professor Murray evaluated Whirl-
wind I in the context of the environment he had seen and
heard interpreted and portrayed by Forrester and his
associates during the Cambridge conversations.
Whirlwind I, they had explained, was to be used primarily
for simulation, but since "no single use of the digital
computer would justify the development cost, " it would
consider two other types of problems: control and
scientific computation. Again Forrester, on this
occasion supported by Everett, evaded typing Whirl-
wind to a particular application. To both men the
question of application was academic, for Whirlwind
was adaptable to a variety of uses, of which ASCA was
only one, even if by contract the primary one.

The report contained no direct criticism of the
Whirlwind program, but Murray did give evidence
supporting the ONR charge of insufficient attention to
the mathematical needs of the program by noting that
no mathematical analysis of the operations of Whirlwind I
had yet been made and any existing plans for one were
inadequate. Such analysis was essential; it should be

5.41

performed within Whirlwind, not by a separate group,
and should be included "as a component of the device. "
There was no need for the mathematical analysis to
await availability of the computer, for it would not
interfere with or delay engineering development. The
two could and should proceed concurrently.

In comparing the two programs at Cambridge and
Princeton, respectively, Murray concluded that although
they had a "common logical ancestry, " they were "distinct
to a remarkable degree. " The application of digital
computation to simultation and control required the
"engineering development" of Whirlwind, a requirement
not imposed upon the IAS computer which was at liberty
to follow "direction of interest to its own objectives, "
namely, the consideration of "purely scientific problems. "
Hence the emphasis upon engineering development was
proper, for engineering development was "absolutely
necessary, " and to delay it would "delay the use of
digital computers in the type of problem" with which

5.42

Whirlwind was concerned. He implied that since
Whirlwind was being designed for future manufacturing,
it had to follow more rigid engineering standards than
did the IAS computer, an implication which von Neumann

• «. a 38
xater rejected.

Once the Murray Report had made its way from

SDC to Mina Rees's office in Washington, a copy was

forwarded to von Neumann at the Institute for Advanced

Studies at Princeton for his comment. Accepting

Murray's definition of Whirlwind' s purpose as "precise

and authentic" and agreeing with the importance of

"a thorough mathematical analysis, " von Neumann

mildly rejected Murray's observations concerning the

differences between the two programs. The contrast

had been drawn too sharply, he felt, yet he rejected

the implication that because Whirlwind had a definite

application in mind and was being designed and

developed with intent of industrial production, it

need be more "reliable and maintainable" than the

5.43

IAS computer which was intended for "general scientific
purposes. "

Von Neumann also questioned Murray's assumption
that the difference in objectives had caused the differences
in design and plan. These resulted, rather, from the
differences in people. If the ojbectives were exchanged,
the courses followed would have remained the same,
for "the subject is new and it is the rule rather than the
exception that two groups who work independently
towards very similar or even identical objectives may
come out with rather different conclusions. I need not
say that I consider this very desirable. The subject is
so new that it is quite reasonable to try a variety of
approaches and not to place all bets on the same
chance.

Von Neumann's observations and judgments were
moderate and restrained and in a vein not unlike Warren
Weaver's of ten months earlier. Unfortunately for
Forrester, they were not strong enough to allay

5.44

suspicions in ONR. Instead, the issue was only just
beginning to be well joined between MIT and ONR, and
it was the sort of issue that many years later was to
provide grounds for the remark, "We're not going to
let it become another Whirlwind!" - a policy view
that could be taken as a stout assertion of control by
a determined administrator or that, again, could be
regarded as a subtle failure of administrative nerve
where the vigorous prosecution of research and
development might be demanded.

NOTES TO CHAPTER 5

1. See End of Chapter 3.

2. Ltr, J. W. Forrester to Director, SDC, Subject:
"Semi-Annual Review of Contract N5ori-60, " p. 2,
January 28, 1947.

3. Ibid. , p. 1.

4. Ibid. , p. 2.

5. Ibid. , p. 1.

6. Ibid. , p. 4.

7. Letter of Intent for Contract NOa(s)-52l6,
December 14, 1944; Task Order No. 1, Contract
N5ori60, June 30, 1945; Amendment No. 4,
January 21, 1948 and Amendment No. 6,
September 29, 1948, Task Order #1, Contract
N5ori60; Directive, Chief of Naval Research to
Director, SDC, February 8, 1949.

8. Interviews by the authors: J. W. Forrester and
R. R. Everett, July 31, 1963, G. S. Brown,
July 6, 1964.

9. J. W. Forrester, Computation Book No. 45,
November 27, 1946 to December 10, 1948,
pp. 89-91 and 134-5, see also the following
memoranda by Perry Crawford, Jr. : Confidential
Memorandum, Subject: "Project Whirlwind, "
November 4, 1947; Memorandum to Director,
SDC, Subject: "Whirlwind Program, " December
18, 1947; Memorandum to Director, SDC, Subject;
"Report on visit to MIT on 9 January 1948, "
January 12, 1948.

10. Memorandum, Perry Crawford, Jr. , to Director,

SDC, Subject: "Report on visit to MIT on 9 January 1948, "
January 12, 1948; Memorandum, Subject: "Report on
conference with Dr. Mina Rees and Dr. John Curtiss
at Sands Point, 15 September 1947, " (author anonymous --

NOTES TO CHAPTER 5 (CONTINUED)

found in ONR files); J. W. Forrester to Capt.
D. P. Tucket, ONR, July 23, 1948; interview,
Norman Taylor by Howard Murphy and
K. C. Redmond, August 8, 1963.

11. Memorandum, Subject: "Report on conference
with Dr. Mina Rees and John Curtis s at Sands
Point, 15 September 1947, "

12. Ltr, J. W. Forrester to Lt. Comdr* H. C. Knutson,
SDD, ORI (Washington), January 28, 1946; Report

L-3, by J. W. Forrester, Hugh R. Boyd, R. R. Everett,
Harris Fahnestock, R. A. Nelson, Subject: "Forecast
for Military Systems using Electronic Digital Computers, "
Servomechanisms Laboratory, MIT, September 17, 1948;
R. R. Everett, The Whirlwind I Computer, revised text
of a paper presented at the joint AIEE- Institute of Radio
Engineers Conference, Philadelphia, Pa. , December
10-12, 1951.

13. Memorandum, Perry Crawford, Jr. to Director, SDC,
Subject: "Discussion of Project Whirlwind at ONR
conference on 28 October 1947, " October 29, 1947.

14. Memorandum, J. W. Forrester to N. McL. Sage,
Subject: "Warren Weaver, Visit to Laboratory,
February 15, 1947," February 19, 1947; J. W.
Forrester, Computation Book No. 45, p. 27; ltr
Warren Weaver to Chief of Naval Research,

att'n, Mina Rees, February 20, 1947; ltr., Warren
Weaver to Chief of Naval Research, att'n Mina Rees,
June 26, 1947.

15. Memorandum for the files, (author anonymous),
August 26, 1947.

16. Memorandum, Subject: "Report on conference with
Dr. Mina Rees and Dr. John Curtis s at Sands Point,
15 September 1947. "

NOTES TO CHAPTER 5 (CONTINUED)

17. I bid. ; Task Order No. 1, Contract N5 ori60,
June 30, 1945; Amendment No. 6, T. O. No. 1,
Contract N5ori60, September 29, 1948.

18. Memorandum, Perry Crawford, Jr. to Director,
SDC, Subject: "Discussion of Project Whirlwind
at ONR conference on 28 October 1947, "
October 29, 1947.

19. Ibid.

20. Project Whirlwind, Summary Report No. 2.

21. Ltr. , J. W. Forrester to Director, SDC, att'n.
Capt. G. M. O'Rear, November 21, 1947.

22. Ltr., J. W. Forrester to H. L. Hazen, Math
Dep't. , March 10, 1947.

23. Ltr. , J. W. Forrester to Director, SDC, att'n.
Capt. G. M. O'Rear, April 23, 1948.

24. Ltr. , Warren Weaver to Chief of Naval Research,
att'n. Mina Rees, February 20, 1947.

25. Ltr. , J. W. Forrester to Director, SDC, att'n
Capt. G. M. O'Rear, November 21, 1947.

26. Memorandum No. 94, W. S. Loud to J. W. Forrester,
R. R. Everett, P. Franklin, Subject: "Suggestions for
Further Work, " August 6, 1947; Memo M-124, Philip
Franklin to J. W. Forrester and R. R. Everett,
Subject: "Location of Target from Combined Observa-
tions, " October 21, 1947.

27. Memo M- 160, P. Franklin to J. W. Forrester,
Subject: "Mathematical Work of Project Whirlwind, "
November 12, 1947.

NOTES TO CHAPTER 5 (CONTINUED)

28. Project Whirlwind, Summary Report No. 3. ,

3 December 1947; ltr. , J. W. Forrester to C. O. ,
Boston Br. Office, ONR, Subject: "Participation
of Mathematicians and Scientists in Project
Whirlwind Activities, " January 30, 1948; J. W.
Forrester Computation Bk No. 45, pp. 86-87.

29. Memo No. M-41, Subject: "Floor Space Required . . . , "
November 29, 1946; ltr. , J. W. Forrester to N. Sage,
November 29, 1946.

30. Ltr. , Jackson and Moreland to N. McL. Sage (copies
to Profs. Forrester and Steves), March 11, 1947.

31. Ltr., J. W. Forrester to Director, SDC, att'n Mr.
H. C. Knutson, April 14, 1947.

32. Ltr. , J. W. Forrester to Director, SDC, att'n
Perry Crawford, April 28, 1947.

33. See: J. W. Forrester, Computation Book No. 44
(October 10, 1946-March 14, 1948); Summary Report
No. 1, to SDC April 1946.

34. Conference Note C-14, Project Whirlwind,
October 9, 1946.

35. J. W. Forrester, Computation Book No. 45, p. 29
(entry of March 17, 1947), p. 52 (entry of August 29, 1947)
and p. 79 (entry of October 20, 1947).

36. See Francis J. Murray, The Theory of Mathematical
Machines. (1947: King's Crown Press, New York.)

37. Memorandum, J. W. Forrester to R. R. Everett,
Subject: "Professor Murray's visit, " November 3, 1947;
memorandum, H. H. Goode, SDC, to Director, SDC,
subject: "Trip to Boston, Mass. , 7-9 November 1947;
Report of, " November 12, 1947.

NOTES TO CHAPTER 5 (CONTINUE!?)

38. F. J. Murray, "Report on Mathematical Aspects
of Whirlwind, " submitted to Director, SDC,
November 21, 1947.

39. Ltr. , John von Neumann to Dr. Mina S. Rees,
December 10, 1947.

Chapter Six
PROBLEMS OF FEDERAL ASSISTANCE

For Project Whirlwind and the Special Devices
Center, 1947 and 1948 were years of increasing
difficulties, even while significant progress in the
design and fabrication of the physical computer
was being accomplished. The joint discharge of the
interfingered administrative and fiscal responsibilities
which the Institute and the Navy bore was complicated
by the organizational and policy changes occurring
within the Navy. In consequence, both Project Whirl-
wind and the Special Devices Center came under inten-
sified administrative and supervisory pressure as the
Office of Naval Research consolidated its responsibility
and authority for certain aspects of the Navy's
research and development program. Misunderstandings
between SDC and ONR — particularly the Boston Branch
Office — led in October to a division of responsibility
for the Project between SDC at Sands Point and the
Boston Branch Office: SDC retained technical super-
vision, but responsibility for "business administration"

1
of the contract was assigned to the Branch Office.

Relations between SDC and ONR continued to deteriorate,

6.01

6.02

nevertheless, until finally technical supervision of

2
the Project also was transferred to ONR. The assump-
tion of direct technical responsibility for Project
Whirlwind was effected by ONR between September of
1948 and February of 1949. It marked acceptance of
the recommendation of its Mathematics Branch that
the Branch "should have the responsibility of promoting
those aspects of the program which involve research,
the dissemination of information, and advising the
Bureaus on novel applications of computers (in systems
or otherwise) which involve research effort." The
Special Devices Center, on the other hand, should be
concerned only with the "application of machines of
proved worth to devices within the scope of their

responsibility, as the computing elements of training

3
devices. "

The transfer of responsibility for Project Whirl-
wind had been a while in the making, but it was inevi-
table, for ONR had consistently demonstrated its deter-
mination to make itself master in its own house. It
could be argued that in formal organizational terms and
perhaps in substantial relationships as well, SDC's
subordinate position was really not inferior to that
which it had held under the Bureau of Aeronautics .
However, ONR was created to perform a mission in the
realm of research and development that existing naval

6.03

bureaus were not to be held responsible for. With
the centralization of responsibility and authority
for naval research and development under ONR, SDC
could not continue to enjoy the wide latitude and
flexibility of operation it had possessed when,
under the aegis of the Bureau of Aeronautics , it
had approved the transition from ASCA to Whirlwind--
that is to say, the transition from a flight trainer
and analyzer to a general-purpose digital computer.
Instead, there occurred a shift in SDC ' s role that
drastically reduced the scope of its activities in
support of research and development.

Although there was a general cut in military
funds for fiscal year 1948, the striking drop in
funds made available to SDC by ONR demonstrates
what had happened to SDC's earlier freedom to select
and sponsor research and development projects. From
approximately eleven million dollars nominally avail-
able to SDC in fiscal year 1947, the amount dropped
to slightly more than a nominal five million for
fiscal year 1948. It was clear by June of 1948
that ONR had lost confidence in SDC's ability to
handle the Project, and furthermore, that ONR was
unwilling to follow Perry Crawford's "fearless and
imaginative jumps into the future" because of
limited funds and because of the belief that "the
present job should be under control before bigger
areas were staked out." Crawford recognized the trend;

6.04

in September of 1948 he accepted a temporary assign-
ment with the Research and Development Board of the
Department of Defense with the intent, upon completion
of the assignment, to return to ONR, but not to SDC.

The delay which ensued between the contractual
change of September, 1948 and the implementing direc-
tive of February, 1949, transferring technical super-
vision to ONR, probably mirrored both SDC's reluctance
to yield completely the traditional freedom of action
it had inherited from its predecessor, the Special
Devices Division, and MIT's reluctance to accept tech-
nical supervision of Project Whirlwind by the Mathe-
matics Division of ONR. By June of 1948, with the
threat of transfer apparently hanging overhead, Nat
Sage became sufficiently concerned to presume to dis-
cuss the threat with Dr. Alan Waterman, the chief
scientist and civilian administrator within ONR. On
this occasion Sage expressed the hope that "in making
any decisions, the Navy would realize the enormous
importance of engineering," including within its
"administrative control . . . persons who understood
the engineering rather than the scientific attack on
a problem." 7

Waterman , grasping the full import of Sage ' s
comments, assured him that the Mathematics Branch
under Mina Rees would not be placed in charge, but
would represent the Navy only on "the mathematical

6.05

8
aspects of the project." Subsequently, in Septem-
ber following, the Head of the Mathematics Branch
was designated "Scientific Officer" for Project
Whirlwind, the title reflecting possibly ONR's ef-
forts to appease both MIT and SDC.

The Special Devices Center continued to exercise
some supervision over the Project until the direc-
tive of February 8, 1949, which assigned "technical
cognizance" to the Mathematics Branch. Thus, between
September, 1948 and February, 1949, SDC had been gra-
dually but completely phased out of the picture. The
contractual amendment of September which had designated
the Head of the Mathematics Branch as "Scientific
Officer" of the Project had also contained the last
allocation to Project Whirlwind out of SDC funds.
The subsequent source of funds became the Physical

Sciences Division of ONR, to which the Mathematics

9
Branch was attached.

The lack of harmony which characterized relations
between Project Whirlwind and ONR since 1947 had ini-
tially stemmed from the apprehensions of the Mathema-
tics Branch over the nature and purpose of the pro-
gram, the quality of its leadership, and its alleged
lack of mathematical competence , on the one hand , and
from the inability of MIT personnel to allay these
apprehensions, on the other.

6.06

An even more profound source of difficulty was
involved, however, even though the differing policy
views with regard to the proper way to go about
developing a computer provided a major source of
friction between Project Whirlwind and ONR. This
more profound source of difficulty lay on the
broadest policy level, from which the federal
government through its agencies in the Executive
and Legislative Branches appraised the operations
of the *cientific community and the value of those
operations to the nation. The pursuit of scientific
research and engineering development to help win the
war usually had been sufficient justification in it-
self for the measures undertaken and the funds spent,
But once the security of the nation was no longer
in daily jeopardy, the conduct of peacetime research
and development (R and D) , together with government
endorsement especially of such R and D, came once
more under renewed , careful , and skeptical scrutiny
by program managers, fiscal officers, and adminis-
trators in the Executive Branch, and by committees
in the Congress . Once more the peacetime policies
of the institution of government toward the estab-
lished institution of science and toward the still
relatively uninstitutionalized activities of an
emerging scientific technology asserted themselves.

6.07

These policies, unfortunately, had been ambiguous and
unsettled at best throughout the nation's history.

When one views the situation from this perspective,
it is not surprising to learn that the proving gap
between Project Whirlwind's vie;^s and those of the
Mathematics Branch brought about, finally, a con-
frontation at the top level between one powerful member
of the private education establishment, MIT, and its
less powerful collaborator-adversary in the federal
military establishment, ONE. This confrontation
occurred at a time when ONR not only was seeking to
assert its responsibility for naval research and
development, but also was striving to gain the con-
fidence of the scientific academic community — a
confidence which, by and large, it ultimately succeeded
in gaining.

The respective positions of MIT and ONR were, in
significant measure , the consequences of the operation
of historical trends transcending either institution's
private history. These trends established the range of
limited freedom of action and the apparent alternatives
allowed to HIT and to the Navy where Project Whirl-
wind was involved. To pause and consider these trends
is to render more understandable and natural, and less
capricious and ''political," the attitudes and actions

6.08

of the principals . It was not an affair that could
be reduced to the appealing, dramatic simplicity of
a sparring match carried on between MIT and Naval
leaders in order to find out who were the more
powerful. Rather, it was one of many smaller events
characterizing the dynamic, historical distribution
and redistribution of judgments and powers continually
taking place and operating to bring about further
adjustments in the subtle, ponderous, and leisurely
process by which human institutions (in this instance,
those of higher education and the national government)
achieve a mutual accomodation over the longer ranges
of time .

Viewed narrowly, it was Project Whirlwind's
relative misfortune to be caught up in this process,
but in the wider arena in which national R and D
policies and practices were at that time being gen-
erated and modified, the stresses to which Project
Whirlwind and ONR both were subjected could well be
regarded as unexceptional. Indeed, had Forrester,
Sage and the others at MIT bowed without a fight to
the pressures which ONR mounted, the affair might well
have been transformed into a "business-as-usual"
situation (although not necessarily a pleasant or
heartening one for any of the parties involved) .

Viewed in a wider perspective , the difficulties
which beset the relationship between Project Whirlwind
and the Office of Naval Research emerge as ramifications

6.09

of the more fundamental difficulties which accompanied
the transition the nation was undergoing as a result
of the Second World War, a transition which was made
even more urgent by the Cold War that set in shortly
after the end of hostilities . It is quite generally
recognized that the Second World War and its after-
math had compelled the nation to abandon its foreign
policy of isolationism and to commit itself to a
role of vigorous, active participation in world
affairs. Less widely appreciated, however, is the
fact that the War also had compelled the American
people and their leaders to reevaluate the role of
science and technology in the national life and to
revise a national posture which in the pre-War years
had been marked by the absence of any popular insistence
that the Federal Government should formulate and
implement a national policy comprehensively to

encourage, coordinate and sustain science and technology

10
as activities of vital concern to the national welfare.

The wartime mobilization and coordination of the

nation's scientific and engineering resources was

neither new nor unique, for previous wars had seen

similar efforts although not as successful or on as

large and authoritative a scale. The continuation of

this pattern in time of peace, however, by the creation

of agencies empowered to direct, coordinate, and fund

6.10

R and D as a substantial and vital part of the national

life was new and without precedent. When the President

on August 1, 1946, signed into law two bills, one

creating the Atomic Energy Commission, the other

the Office of Naval Research, tangible proof was

offered that the government had accepted and was

implementing the principle of a continuing and

comprehensive responsibility for the advancement of

science and technology. The subsequent establishment

of similar agencies, such as the Air Research and

Development Command of the Air Force, the National

Science Foundation, and the National Aeronautics and

Space Administration, give further evidence of this

continuing acceptance .

The statutory creation of the Office of Naval

Research marked a victory within the Navy Department for

a group of dedicated and perceptive civilians and

military officers who early in the War had seen the

need for the creation of a central office to coordinate

and direct naval research and development. In the

pre-War years naval R and D had been uncoordinated and

routine. Conducted to as great an extent as possible

within the laboratories of the various Bureaus and

the Naval Research Laboratory, it had been concerned

primarily with the improvement of existing procedures

11
and equipment. The exception to the routine nature

6.11

of naval research and development was that conducted
under the aegis of the Naval Research Laboratory where
a taste for fundamental work had been developed. The
Naval Research Laboratory was, however, "a very small

exception to the general lack of research in both

12
Army and Navy . "

During the War years as the government ' s total
expenditures for R and D mounted, so did the Navy's.
The latter' s costs rose from $13,566,899 in fiscal
year 1940 to $149,887,877 in fiscal year 1944. Total
expenditures for this five year period, including
monies transferred to other agencies, approximated
$405,000,000, about twenty-two percent of the govern-
ment's total expenditures for the period. Of the
$405,000,000, the Navy disbursed $348,626,000 itself:
$97,853,000 in its own laboratories; $248,834,000 to
private industrial laboratories; and $1,939,000 to
education and foundation laboratories. The rising
trend established during the war years was, with
minor and occasional cutbacks, to be carried over
into the post-war period.

Throughout the War, the Navy's bureaus continued
to bear primary responsibility for research and dev-
elopment within their respective areas of responsibility.
Coordination of the bureaus ' respective programs was
attempted, however, by the creation in the Office of
the Secretary of the Navy of the Office of the

6.12

Coordinator of Research and Development. In addition
to coordinating internal research and development
programs , the Office of the Coordinator represented the
Navy on the boards of other research agencies, main-
tained liaison with the Office of Scientific Research
and Development, and kept the Secretary and the
various bureaus and offices informed of research

and development programs both within and without the

14
Navy.

The Coordinator continued to provide whatever
central direction there was to the Navy's research
and development programs, until the impending dissolution
of OSRD, together with the growing complexities of
the programs , caused the Secretary of the Navy to
establish a central agency, the Office of Research
and Inventions. This agency embodied in itself
the Office of the Coordinator of Research and Develop-
ment , the Office of Patents and Inventions , the
Naval Research Laboratory, and the Special Devices
Division. The aims of the new ORI were to (1)
stimulate research and development throughout the
materiel bureaus, (2) assume cognizance where a
project was of major interest to more than one bureau,
and (3) undertake by contract or within its own

laboratories "fundamental work not unique to any

15
single bureau." The creation of the Office of

Research and Inventions indicated that the Navy was

6.13

laying plans for the post-war period in recognition of
the necessity to coordinate and direct the research
and development programs of the respective bureaus
and also aggressively to pursue fundamental research
in areas pertinent to naval science and technology.
Subsequently, these functions of the Office of
Research and Inventions were given Congressional
sanction when the Congress established the Office of
Naval Research.

The Office of Naval Research represented a
substantial victory for proponents of the principle
of a central departmental authority to coordinate
and direct Naval research and development, including
basic research. The establishment of the Office of
the Coordinator of Research and Development had been
a major step in this direction and one accomplished
over the protests of the General Board. The subsequent
creation of the Office of Research and Inventions had
advanced the principle considerably by providing it
with Presidential sanction as well as Secretarial.
The legislative establishment of the Office of Naval
Research added Congressional acceptance of the principle
to Presidential and Secretarial. Most importantly,
Congressional approval guaranteed appropriation of
the funds necessary to fulfillment of the principle —
a guarantee not implied in the sanctioning of the

6.14

Office of Research and Inventions by Executive Order.
Indeed, the Navy had been compelled to request of
Congress the passage of legislation establishing the
Office of Naval Research, because the House Appro-
priations Committee had refused to consider monies

for the Office of Research and Inventions until the

1 K
approval of the Congress had been given.

Congressional resistance to the Office of

Research and Inventions did not imply objections to

the principle of centralization; rather it reflected

the resistance of Chairman Carl Vinson of the House

Committee on Naval Affairs to continuing Executive

"use of war powers in peacetime." Such use, Vinson

had warned, "could seriously impair the relations of

the Navy Department with the Congress." In fact, the

Committee on Naval Affairs had strengthened the Office

by writing into the bill changes which gave the Office

"control of all naval research," including — subject

to Secretarial approval — authority to control the

research programs of the bureaus . Thus , if the

Secretary approved, the Office of Naval Research

presumably would exercise authority over the total

17
spectrum of Naval research and development.

The Vinson Bill creating the Office of Naval
Research became law on August 1, 1946, thus ante-
dating by some four years the creation of the altered
peacetime successor to the Office of Scientific

6.15

Research and Development, the National Science Found-
ation. Consequently, the Navy was one of the first
government agencies to fill, at least partly, the
void created by the phasing out at the War's end of
the Office of Scientific Research and Development.
In fact , the imminent end of the War had hastened the
administrative and legislative steps which culminated
in the establishment of the Office of Research and
Inventions and its successor, the Office of Naval
Research.

From the beginning the primary task of the Office
was the sponsorship of basic research. Development
was to remain with the respective bureaus. The first
Chief of Naval Research, Vice Admiral Harold G. Bowen,
implemented this policy by supporting research of the
most fundamental nature, and by 1947 the Office had
planned a research program which would cost about
$20,000,000 annually. Toward the end of 1948 the
Office had in its employ some 1,000 scientists distri-
buted among three in-house laboratories and six branch
offices. It had contracted for some 1131 projects at
200 institutions, a program accounting for approxi-
mately forty percent of the nation's total program
in basic research, according to one estimate. The
value of contracts in which ONR was involved approxi-
mated $43,000,000, of which $20,000,000 came from
the Office's own funds; $9,000,000 came from other

6.16

federal agencies — principally the Atomic Energy
Commission — but distributed by the Office of Naval
Research; and $14,000,000 came from various univer-
sities, according to one tally. Thus, during the
crucial post-war years while the Congress was de-
bating the kind of organization which should be
created at the national level to sponsor basic
research, the Office of Naval Research was actively

pre-empting the field and continuing a program that

19
many considered vital to the national security.

The founders of the Office of Naval Research had
assumed that the Office, once established, would
•mount a comprehensive and sustained program in basic
research, and one not restricted to areas of Naval
pertinence only. The proponents of a peacetime
Office of Scientific Research and Development , how-
ever, had assumed that Navy responsibility for a
comprehensive program would be temporary, pending
Congressional authorization of a national agency
purposed to sponsor basic research in its broadest
sense. Vannevar Bush, one of the most forceful
advocates of a comprehensive national agency, had
supported the Navy undertaking while voicing the
reservation that it was "entered into with the full
understanding on the part of everyone that it was to
a considerable extent a temporary program, and that
if the Congress saw fit to establish a Foundation for

6.17

the purpose, the principal burden of that work would
be transferred to the Foundation."

The Navy would continue to sponsor some projects,
Bush opined, but the bulk of basic research would be
the responsibility of the foundation, where it could
be managed "by a group which can combine the military
and the civilian points of view and which can judge
the thing from a somewhat broader basis than the ser-
vices, by their very nature, can hope to judge it."
Despite Bush's reference to an initial "full under-
standing," the Office of Naval Research did not
share his opinion. While it was willing to transfer
some Navy projects of broad interest to the proposed
agency, it intended to continue its own program at
the fiscal level already attained, expanding in areas
of immediate pertinence.

The expansive powers granted the Office of Naval
Research by the Congress implied the eventual central-
ization of Navy-sponsored research under the new office,
As ONR sought to implement the authority inherent with-
in its enabling legislation, it took steps that had
profound consequences for the Special Devices Division
and for Project Whirlwind. Within a span of two years,
as has been noted, the Special Devices Division was
subordinated as the Special Devices Center, and Project
Whirlwind was transferred to the jurisdiction of the
Mathematics Branch of the Office of Naval Research.
For Project Whirlwind this meant that the sympathetic ,

6.18

understanding, fiscal supervision and program encour-
agement of the engineers of the Special Devices
Center was replaced by the skeptical, less-than-
enthusiastic supervision of the mathematicians of
the Mathematics Branch. It did not help matters
that the latter were more interested in the computer
as a tool for scientific computation than as the
"brain" of a command-and-control center for tacti-
cal and logistical operations , such as envisioned
by Forrester, Everett, Perry Crawford, and others

at Sands Point who became familiar with "L-Notes"

21

L-l and L-2 which the two MIT engineers had written.

Another complicating factor in the relations
between the two groups was the rising cost of carry-
ing forward Project Whirlwind at the very time the
nation was undergoing post-war military retrenchment,
with its impact upon military budgets. Funding, if a
problem, had been a very minor one with no discernible
effect until the fall of 1948. From then the question
of money was to overshadow all others and continue a
chronic source of irritation and difficulty.

The initial amount of money — $1,194,420 — committed
by the Navy to the Project under the terms of Task
Order No. 1 of Contract N5ori-60 had been increased
two years later by an additional $100,000 and again

Oh

in January of 1948 by $520,000. The first increase
was required apparently to meet the extra costs incurred

6.19

by extension of the contract's terminal date for
one year to June 30, 1949. The second increase was
intended to defray the costs of program acceleration
requested early in 1947 by SDC. 25 In addition to
expanding and expediting the program at MIT, part of
the work was subcontracted to Sylvania Electric
Products , Incorporated at an estimated cost of
$319,576.75. Sylvania was to "conduct studies and
experimental investigations in connection with:
'final packaging design and construction of the

9 R

Whirlwind I electronic digital computer.'"

The cost of the Sylvania subcontract was in-
cluded in the request for additional funds for
fiscal year 1948 which was forwarded to SDC, in
August, 1947, totalling some $441,520.75. These
funds, it was noted, would not cover the costs "for
photographic input-output devices" to be purchased

from Eastman Kodak or for "aircraft simulation com-

27

ponents" which ONR would have to fund separately.

The amount finally allocated by the Navy was $520,000,
approximately $80,000 above the MIT request, but be-
tween the time of the request and final Navy action,
Project Whirlwind had entered into a subcontract
with Eastman Kodak in the amount of $70,000 for
"photographic storage equipment . . . necessary
to the completed simulator." 28 Neither MIT's
request nor the Navy's final allocation caused any

6.20

furor in either organization. Jay Forrester did
comment to Nat Sage that the estimated cost con-
tained in Amendment Number 4, which officially

allocated the additional funds, was some $500,000

29
short, but there the question died.

The "blowup" came in the fall of 1948, fol-
lowing a letter from Nat Sage to SDC in which he
requested for the Project funds in the amount of
$1,831,583 for the fifteen-month period between
July 1, 1948 and September 30, 1949. This amount,
when added to an unexpended amount of $385,260,
produced a total of $2,216,843 or a monthly expendi-
ture rate of approximately $150,000. This figure
raised havoc with ONR's budget and brought into the
problem both the Chief of Naval Research and MIT's
top administration.

Earlier, MIT's top management had become con-
cerned about the friction which had developed be-
tween Project Whirlwind and ONR, and presumably had
become at first a trifle- dubious either about its
ignorance of the Project in detail or about the
Project itself and its management. To determine the
quality of the program and its leaders , the Institute
leadership through Dr. James R. Killian, Jr., then
vice-president, asked Ralph Booth, a member of the
MIT Corporation's Electrical Engineering Committee,
to review the status of the Project. This was in

6.21

the winter and early spring of 1948. Booth, ques-
tioning his own competence to "pass on the theoret-
ical and technical merits" of the electrostatic
storage tube under development, retained as a con-
sultant Dr. J. Curry Street of the Harvard physics
faculty and formerly a member of the MIT Radiation
Laboratory .

Booth, and presumably Street, visited the
Project in May, July, and August of 19 4 8 to study
its operations and obtain the information and im-
pressions which would be necessary for an appraisal
of its worth. Booth submitted his report to Killian
on August 26, prior to the exchange of views which
took place between ONR's and MIT's top administrators
in September and December of 1948.

In his report to the vice-president, Booth
stated that the purpose of his review had been "to
determine whether the accomplishments to date and
the organization and procedure of the work currently
in hand insured a successful completion of the pro-
ject approximately in accordance with the present
schedule." His report was very favorable. Presumably
supported by Street, Booth observed that the Project's
"Accomplishments . . . give every promise of providing
within the scheduled date a successful computer at
speeds hithertofore unrealized." With the exception
of the storage tube, the program in all its phases

6.22

had reached that point , he noted , "where the remaining
work can be classified as design engineering or devel-
opment of refinements." The storage tube, he antic-
ipated, would be successful; however, if not, the
other components of the computer could be adapted to
other types of memories with no greater penalty than
some loss of speed in computation.

Booth and Street were so enthusiastic about the
potential of the storage tube that they strongly urged
the "Navy be asked to acquaint itself with the high
promise of this development, since it is entirely
possible that this tube may supplant mechanisms which
hold less promise and which are in an earlier stage
of development and on which appreciable sums of Navy
research money are currently being expended." All
in all, Booth, found the Project to be "well-organized,

staffed by efficient, capable people, and . . . con-

30
ducted in proper accord with the timetable . . . "

It is not surprising to learn that when Booth's
laudatory comments were added to the support rendered
the Project by Nat Sage and Gordon Brown, the MIT
top administrators considered themselves sufficiently
well informed to become convinced that Project Whirl-
wind was worthy of their support. So they came to
its rescue in the funding crisis of 1948.

Nat Sage's request on August 4 for 1.8 million
dollars to cover costs through fiscal year 1949

6.23

and the first three months of fiscal year 1950 may
not have caught the Navy totally unawares, but it
was nevertheless an irritating if not downright
disturbing request. Not only did it ask ONR to
double the amount ONR had already committed to the
Project, but it was submitted some thirty days after
the fiscal year to which it was to apply (FY 1949)
had begun. To the Navy it must have been a splendid
example of the continuing, erratic and unpredictable
pattern of behavior followed by Forrester and his
colleagues who, between the spring of 1947 and the
fall of 1948, had raised their estimates of financial
requirements for fiscal year 1949 by some 1.3 million
dollars. Here was a pattern particularly disturbing

to administrators whose policies were controlled more

31
by financial considerations than by technical.

Looking back, one could see that in the spring
of 1947 both MIT and SDC had agreed that fiscal year
1949 costs might equal a half million dollars. But
by October, 1947, MIT foresaw fiscal year 1949 costs
of $940,000, and in December it raised the estimate
again to 1.2 million dollars. Since the fiscal year
would begin July 1, 1948, it was then a bare six
months in the offing.

ONR, in the meantime, had been raising its esti-
mates at a different pace . Although ONR ' s subordinate ,
SDC, had agreed on the half-million-dollar figure in

6.24

the spring of 1947, its reaction to MIT's $940,000
estimate was to go up only to $600,000. But during
the spring of 1948, SDC once again came into agree-
ment with MIT, now at the new, higher level of 1.2
million. Then in June of 1948, just before fiscal
year 1949 was to begin, Admiral Paul F. Lee, Chief
of Naval Research, cut ONR's support back to $900, 000. 32
The following month, after the beginning of fiscal
year 1949 , Forrester proposed to raise the ante to
1.8 million dollars for the fiscal year which had
already begun. ONR's response by Admiral Lee's
successor, Admiral T. A. Solberg, was a courteous
but firm "no"; $9 00,000 would remain ONR's commitment.

Admiral Lee's reasons underlying his decision
to reduce Project Whirlwind's allocation to $900,000
for fiscal year 1949 were presumably many and complex,
but a lack of funds was not included among them. At
the time Lee made his decision it is reasonable to
assume that he must have had a fairly good idea of
what the amount of unexpended monies to be carried
over from fiscal year 19 48 would be, for at the time
the 1949 budget was under consideration, such monies
were estimated to approximate $32,000,000.

Other considerations had to provide his guide-
lines, therefore. In view of the pattern of events
involving the transfer of technical responsibility
for Whirlwind from SDC to the Mathematics Branch,

6.25

which was to take place during the winter of 19 48
then approaching , one of the consideration* very likely
was the intent to bring SDC into its proper relation-
ship with the rest of the ONR structure — in this case
by diminishing SDC's role in computer research. The
decision had been made only after a careful review
of SDC's programs for fiscal year 1949 33 and the
$900,000 permitted SDC for allocation to Project
Whirlwind represented the last monies SDC was to
receive for this purpose. Future funds were to be
allocated and controlled first by the Physical
Sciences Division, to which the Mathematics Branch
was attached, and subsequently > after its formation,
by the Mathematical Sciences Division.

6.26

Furthermore , the amendment which announced the
allocation also announced the appointment of the Head
of the Mathematics Branch as "Scientific Officer" of

Oh

the Project, ^ and within six months Lee's successor
was to eliminate SDC from the program completely by

giving direct control of the Project to the Mathe-

35
matics Branch. As has been noted above, Perry

Crawford had recognized the trend and left SDC.

Another primary reason, presumably, was Lee's
intent to bring Project Whirlwind under firm control.
There is no sign that he was discouraged in this by
the Mathematics Branch. On the contrary, the Branch
had been concerned and disturbed about the Project
ever since ONR had become responsible for it. The
Project by its own behavior had provided some sub-
stance to feed the fears of the Mathematics Branch.
Within a period of eighteen months, at most, Forrester's
estimates of the additional financial needs for fis-
cal year 1949 had escalated from $500,000 to $940,000
to $1,200,000 to the final figure $1,831,583, and
this final figure exceeded by a magnitude of three
the combined allocations for the two previous fis-
cal years.

For that matter, it exceeded by some $700,000 the
original monies obligated under the terms of the con-
tract when first negotiated. Moreover, the 1.8 31 mil-
lion (1.465 for the 12-month fiscal year) almost matched

6.27

the 1.850 million which Lee, in his testimony sup-
porting the proposed Navy budget for fiscal year 1949,
had estimated would be obligated in the entire gen-
eral area of mathematical research. The Whirlwind
request, which alone would have used approximately
ten percent of ONR's 1949 funds for contract research,

thus threatened to consume almost the total amount

3 G
designated for research in mathematics.

It is to be hoped that Admiral Lee understood
his job well enough to have considered both the im-
pact of his reduction upon MIT alone and its rever-
beratory. fiscal effect upon the rest of the academic
community with which ONR dealt. And this hope is
borne out by further consideration of the fact that
the Navy, through its various bureaus as well as ONR,
maintained active research and development programs

in the nation's universities, spending some $25,000,000

37
in fiscal year 1948 alone. In thxs connection

it is doubtful that the 1948 funding crisis of 1.8

million dollars with MIT would have had drastic

financial effect upon university relations in general.

However, the psychological impact might have been

considerable, particularly upon the fragile, newly

established relations between ONR and the universities;

in this respect the matter had to be handled with the

greatest finesse and diplomacy.

Whatever Lee's reasons, Project Whirlwind's

6.28

requested allocation for fiscal year 1949 had been
cut almost in half, and the monies would be some
$550,000 less than had been planned for. This threat-
ened a reduction in the planned monthly expenditures,
from a rate of $150,000 to an amount slightly in excess

of $10 5,000, if the group were to stay within the finan-

3 8
cial limits imposed by ONR's decision. It was a pro-
posed cutback of approximately thirty percent, and its

repercussions would be severe. The program's rate
of progress as planned by the MIT group and accepted
in the main by SDC, although not by ONR, would be
seriously curtailed and at a time when Forrester and
his colleagues were quite sanguine in their belief
they were on the edge of success. The reaction of
Forrester and Sage to the threat of inadequate funds
to meet the planned schedule provoked a major issue
that was then carried to the highest administration
levels of both ONR and MIT.

On September 2, 1948, the new Chief of Naval
Research, Admiral T. A. Solberg, approached the Pres-
ident of MIT, Dr. Karl T. Compton, with the suggestion
that in light of the wide discrepancy between the funds
requested by Project Whirlwind and the allocation made
by ONR for fiscal year 1949, "future commitments and
rate of expenditure be scaled down," pending an evalu-
ation "of both the technical and financial requirements

6.29

of the project." Such an evaluation, he implied,
might result from a study of all computer programs
then being conducted by the Computer Sub-Panel of
the Research and Development Board. In the meantime,
he proposed a conference be called between MIT and ONR
to "reexamine both the technical and financial scope
of the project" in order to clarify "future policy

and . . . establish a firm basis upon which Project

39
WHIRLWIND should operate in the future."

Solberg ' s letter to Compton brought the MIT
president directly into the matter and precipitated
within Project Whirlwind a flurry of activity aimed
toward indoctrinating Compton and winning his support
to the Project. Within a few days after receipt of
the letter, Compton conferred on the matter with
James R. Killian, Jr., his vice president (very shortly
to succeed him), with Nat Sage, with F. L. Foster,
Sage's assistant in the Division of Industrial Coopera-
tion, and with Jay Forrester. At this conference
Forrester presented to the group his appraisal of
"the size of the total digital computer program . . .
the United States was facing," estimating at that time
that costs would run some $100,000,000 per year for
ten years "if the apparatus that people counted on
getting was to be made available." In his comments,
Forrester included the opinion that some estimates of
time and cost tended "to be hundreds of times too low."

6.30

The MIT president was apparently enough impressed
by Forrester's presentation to request that it be
suitably prepared to permit him to take it to
Washington for distribution within influential
official circles. This informal request was followed
on the same day by a formal request to Sage from
Compton, in his capacity as president of MIT and
hence ultimately responsible for Project Whirlwind and
also in his capacity as an advisor to the Armed
Services, to prepare a report. This report would
clearly present the "potentialities for useful appli-
cations inherent" in the digital computer and would
give some estimate regarding the "time , money and
staff" which would be necessary to "carry digital
computing equipment to the point of use by the Armed
Services . "

Such a report, Compton noted, would not only
be of immense help to him as he sought to grasp
fully the potential use and cost of digital computer
programs in general and Whirlwind in particular,
but would be also of great benefit to ONR and to any
other organizations which might be considering the use
of the digital computer. It obviously would provide
Forrester with an excellent opportunity to arrange
his thoughts and to gain access for them through
Compton to higher command levels within the Government,
an important factor in the struggle with ONR which

6.31

was looming on the horizon.

The decision was also taken at the internal MIT
conference to press Ralph Booth for his formal eval-
uation of the Project. (The conferees could be sure
it would be favorable , in light of his complimentary
comments in his letter of the previous August 26 to
Killian.) It was further decided to ask Booth to
serve as a representative of MIT in the forthcoming
meeting with the Chief of Naval Research and his
staff. Apparently their own background knowledge,
Booth's letter of endorsement, Forrester's presen-
tation, and Sage's judgment and confidence in the
Project persuaded Compton and Killian that the Pro-
ject was in competent hands and had a significant
contribution to make.

Although the record does not show it, the
Institute's leaders may also have recognized that
here was a test case made to order upon which they
could make a stand suitable to the purposes and need
of establishing viable practices and durable relation-
ships favorable to the continuing conduct of military-
sponsored research by private universities. It is
not unreasonable to suppose that those responsible
for Institute policies — who were already involved in
an emerging, loose, but effective organization of
civilian scientists (known informally among scientists
two decades later as the "Eastern Establishment") ,

6.32

the aim of which was to maintain the intelligent
prosecution of private scientific and engineering
research funded by federal interests—were astute
enough to realize that three years had passed since
the War had ended, that the shakedown period into
peacetime procedures was drawing to a close, causing
these procedures to lose the plastic flexibility
they had possessed when new, that the tenor of inter-
national affairs was becoming increasingly discordant
as a consequence of Stalin's vigorous intransigeance ,
aid that the computer technology then dawning offered
prospects and applications in war and peace that
quite transcended those afforded by the usual mili-
tary research project. In any event, whether they
were moved or not by such explicit long-range con-
siderations in addition to their informed faith in
the competence of Project Whirlwind, the MIT leader-
ship made elaborate preparations that beggared those
undertaken in ONR.

The confidence and the support engendered at
the Institute were displayed not only by Compton's
request to Sage for the report he wished to circu-
late in Washington, but also by Sage's observation
that "these reports must be gotten into various
people's hands fairly promptly." Even more emphat-
ically, it was demonstrated by the men Compton
appointed to represent MIT in the forthcoming meeting

6,33

with ONR, for once the Institute's position had been
determined and Solberg's proposal accepted, Compton
nominated Jay Forrester and Nat Sage, along with Ralph
Booth, to argue the Project's cause — three men whose
views were known, whose biases and commitments in the

matter were shared, and whose policy views were in

"+1
close accord with those of MIT's top management.

The Institute leadership had heard the case, had

rendered its judgment, and had not found the Project

wanting. Thus prepared, they were ready to meet

with ONR.

NOTES TO CHAPTER 6.

1. Memorandum, Perry Crawford to Director, SDC, subj . :
"Trip to Boston, Mass., 8 January 1947 to 10 January
1947; report of, "January 28, 1947; memorandum for the
file [no author], subj.: "SDC Computer Section Com-
ments on Reference (a)," August 26, 1947 [from SDC's
files]; ltr., W. R. Mangis, to Director, SDC, and to
CO, Boston Branch Office, ONR, subj.: "Contract N5ori-
60 (M. I. T.) - Assignment of Contract Administration
Responsibilities," October 2, 1947.

2. Memorandum, Perry Crawford to Director, SDC, subj.:
"Report on visit to MIT on 9 January 1948," January
12, 1948; Amendment #6, T. 0. #1, Contract N5ori60,
September 29, 1948; ltr., T. A. Solberg, Chief of
Naval Research, to Director, SDC, subj.: "Contract
N5ori60 Task Order I Massachusetts Institute of
Technology; change in cognizance of," February 8, 1949.

3. Memorandum, N424 (Mina Rees) to N101, subj.: "Responsi-
bility for Computer Research and Development," January

28, 1949.

4 . Annual Report of the Secretary of the Navy for the Fiscal
Year 1947 , Washington, 1948, p. 82.

5. Ltr., N. McL. Sage to Jay Forrester, June 10, 1948.

6. J. W. Forrester, Computation Book No. 45, November 27,
1946 to December 10, 1948, pp. 134 and 151.

7. Ltr., N. McL. Sage to Jay Forrester, June 10, 1948.

8. Ibid .

9. Amendment #6, T. 0. #1, Contract N5ori60, September

29, 1948; Ltr., T. A. Solberg, Chief of Naval Research,
to Director, SDC, subj.: "Contract N50ri-60 Task Order^

I Massachusetts Institute of Technology; change in cogni-
zance of," February 8, 1949; Amendment #7, T. 0. #1,
Contract N5ori60, March 31, 1949.

10. See Kent C. Redmond, "World War II, a Watershed in the
Role of the National Government in the Advancement of
Science and Technology," Charles Angoff (ed.), The Humanities
in the Age of Science (Rutherford, N. J., 1968), pp. 166-
180. For an excellent history of the federal government
and science, see A. Hunter Dupree, Science in the Federal

Government : A History of Policies and Activities to
1940 (Cambridge, Mass., 1957) .

11. Annual Report of the Secretary of the Navy - Fiscal
Year 1910 , pp. 25-7; see also the Report for FY 1939,
pp. 24-5.

12. Dupree , Science in the Federal Government , p. 333.

13. The Government's Wartime Research and Development,
1940-44; Report from the Subcommittee on War Mobiliza-
tion to the Committee on Military Affairs pursuant

to S. Res. 107 (78th Congress) and S. Res. 146 (79th
Congress) Authorizing a study of War mobilization
problems, July, 1945, Part II: "Findings and Recom-
mendations," 79th Cong., 1st Sess., Senate Subcom-
mittee Report No. 5, pp. 5, 56-67, 70-71. Senate
Hearings before a Subcommittee of the Committee on
Government Operations, House of Representatives, 85th
Cong., 2nd Sess., July 14, 15, 17, and 18, 1958, on:
Research and Development (Part 2 — Military Research
Representatives) .

14. Annual Report of the Secretary of the Navy for the
Fiscal Year 19~4T , "Washington, r9~4~2T~P-~TI7 Tfie
Bird Dogs, The Evolution of the Office of Naval
Research," Physics Today , XIV (August, 1961) 31-2.

15. Committee on Naval Affairs, House of Representa-
tives, Hearing on H. R. 5911 (ex-4317), to Establish
an Office of Naval Research in the Department of the
Navy, March 26, 1946, pp. 2821-22. Annual Report,
Fiscal Year, 1945, The Secretary of the Navy to the
President of the United States, pp. 30-1; see also
the Secretary's Annual Report for 1946, pp. 66-73.

16. Hearing ... to Establish an Office of Naval Research,
pp. 2834-7.

17. Ibid ., pp. 2840-57.

18. '''Evolution of ONR," Physics Today , XIV (August, 1961)

33-5.

19. Hearing ... to establish an Office of Naval Research,

p. 2 847; John E. Pfeiffer, "The Office of Naval Research,"
Scientific American , 180 (February, 1949) 11-5; Carroll
W. Pursell, Jr., "Science and Government Agencies," David
D. Van Tassel and Michael G. Hall (eds.), Science and
Society in the United States , (Homewood, 111., 1966),
pp. 245-6,

20. Hearings before the Committee on Interstate and Foreign
Commerce, House of Representatives, 80th Cong. 1st Sess.
on H. R. 942, H. R. 1815, H. R. 1830, H. R. 1834, and

H. R. 2027, Bills relating to the National Science
Foundation, March 6 and .7, 1947, pp. 208, 231-54.

21. Task Order No. 1 to contract N5ori-60, June 30, 1945,
and amendments No. 4, January 21, 1948; No. 6, September
29, 1948; and No. 9, July 1, 1949. Interview, J. W.
Forrester and R. R. Everett by the authors, July 24, 1964

22. Task Order No. 1, Contract N5ori-60, June 6, 1945.

23. Amendment No. 3, T. 0. No. 1, Contract N5ori-60, June
26, 1947.

24. Amendment No. 4, T. 0. No. 1, Contract N5ori-60, January
21, 1948.

25. J. W. Forrester, Computation Book No. 45, p. 35.

26. Ltr., (illegible signature) to MIT, DIC, subj: "Con-
tract N5ori-60, Task Order 1, Proposed Cost-plus-fixed-
fee Sub-contract to Sylvania Electric Products, Inc.,
in the aggregate amount of $319,576.75," September 22,
1947; Prime Contract No. N5ori-60, Sub-contract No. 1,
DIC Project No. 6345, Revision No. 1, September 24, 1947.

27. Enclosure "A", "Proposed form of letter for budget
request to Office of Naval Research," to ltr., J. W.
Forrester to F. L. Foster, DIC, MIT, August 14, 1947;
ltr., F. L. Foster to CO, Boston Branch Office, ONR,
subj.: "Contract N5ori-6 0, Budget for Fiscal Year
1947-48," August 15, 1947; Memorandum for Files, J.
B. Thaler, Procurement Officer, SDC, subj.: "Status
of Contract N5ori-60 with Massachusetts Institute of
Technology and N6ori-133 with McKiernan Terry Corp.
Summarization of facts and events leading thereto,"
October 23, 1947.

28. Memorandum, J. B. Thaler, SDC, to H. W. Fitzpatrick,
Chief Accountant, ONR, subj.: "Contract N5ori-6 0,
Task Order 1, Massachusetts Institute of Technology;
Financing thereof," November 3, 1947.

29. Ltr., J. W. Forrester to Nat Sage, subj.: "Amendment
No. 4 to Project Whirlwind Contract N5ori60," February
2, 1948.

30. Ltr., Ralph D. Booth to Dr. James R. Killian, Jr., Vice
President, MIT, August 26, 1948.

31. Memorandum, Perry Crawford, Jr. to Technical Director,
SDC, subj . : "Project Whirlwind, Fiscal Requirements
of," July 22, 1948.

32. Memorandum, C. H. Doersam, Jr. to Perry Crawford,

subj.: "Project Whirlwind 24-x-3, Budget Estimate

of," October 24, 1947; memorandum, Perry Crawford, Jr.
to Technical Director, SDC, subj.: "Project Whirlwind,
Fiscal Requirements of," July 22, 1948; Itr., J. R.
Ruhsenberger, SDC, to Chief of Naval Research, att'n.
Code N101, subj.: Project Whirlwind, Contract N5ori-60,
Financing of," August 18, 1948.

33. Ltr., J. R. Ruhsenberger to Chief, ONR, August 18, 19 48.

34. Amendment No. 6.

35. Ltr., Solberg to D. R. , SDC.

36. Hearings before the Subcommittee of the Committee on
Appropriations, House of Representative, 80th Congress,
2nd Session, on the Department of the Navy, Appropriations
Bill for 19 49, p. 96 8; Amendment #6, T. 0. #1, Contract
N5ori-60, September 29, 1948; Ltr., J. R. Ruhsenberger,
SDC, to Chief of Naval Research, att'n. Code N101, subj.:
"Project Whirlwind, Contract N5ori-60, Financing of,"
August 19, 1948.

37. Hearings on S. B. 1560, p. 28, March 1948.

38. Ltr., J. R. Ruhsenberger, SDC, to Chief of Naval Research,
att'n. Code N101, subj:: "Project Whirlwind, Contract
N5ori-60, Financing of," August 19, 1948; Procurement
Directive, Contract N5ori-60, T. 0. #1, Directive #Nr-
720-003/7-22-48, August 11, 1948, attachment "C" Clearance
Memorandum prepared by J . B . Thaler , Procurement Officer ,
SDC.

39. Ltr., T. A. Solberg to K. T. Compton, September 2, 1948.

40. N. McL. Sage, "MEMORANDUM of Conference between Dr. Compton,
and Messrs. Killian, Forrester, Foster and Sage on Project
Whirlwind," September 8, 1948; ltr., K. T. Compton to

N. McL. Sage, September 8, 1948.

41. N. McL. Sage, "Memo on Conference ... on Project Whirl-
wind," Sept. 8, 1948; ltr., R. D. Booth to Dr. J. R.
Killian, Jr., Aug. 26, 1948; ltr., Henry Loomis to T. A.
Solberg, Sept. 10, 1948.

Chapter Seven
BREAKING NEW TRAILS

Project Whirlwind obtained the support of the
Institute leadership in part because of the informa-
tion and attitudes and judgments that Nat Sage and
Jay Forrester conveyed and in part because of the
engineering operation that Forrester and his asso-
ciates had been mounting in the Barta Building.

Within the Project, under Robert Everett's
leadership during 1947 the operating requirements
of the proposed computer had been incorporated into
"block diagrams" stipulating the coordinated and
systematic operation of the basic functional com-
ponents of the proposed machine. Using the block
diagrams as master plans specifying the performance
of the components singly and together, Everett,
Forrester and several of the engineers then pro-
ceeded during 1947 and 1948 to lay out and review
the design of appropriate electronic circuits. These
would carry out the physical operations which would
correspond to the mathematical and logical operations
associated with binary digital computation and with

7.1

7.2
the storing, retrieving, and evaluating of such
digital information.

Since this is not an engineering history of
the Whirlwind machine that was designed, built, and
put into operation between 1947 and 1951, specific
detailed analyses of the many engineering problems
encountered and the solutions worked out have no
place here . In the view of the authors , the inside
engineering story available only to readers possessing
a specialized scientific and engineering technical
education does not provide the only means of obtaining
an illuminated understanding of the research and
development process under case study here -- a pro-
cess which has become a characteristic social and
economic activity of twentieth-century America.
Clearly, the technical engineering progress accom-
plished by the Project Whirlwind engineers continually
influenced the course of events, and equally clearly,
the engineering story vitally affected the eventual
outcome of the enterprise. To give these aspects
of the larger story the justice that is their due,
a technical digression would be required that is
beyond the scope of this case study. Consequently,
in selecting an alternative to the specialist's
route to understanding, the authors sought to convey
the import and the general character of the engineering

7.3
activity of the Project by indicating in the lan-
guage common to us all the more important events
that occurred in the uncommon -sense realm of science
and engineering.

In general terms, the young MIT graduates in
charge of the enterprise faced the task of converting
mathematical, logical, abstract concepts into working
machinery. The abstract models they conceived and
worked up began, for the most part, with theoretical
considerations of the arithmetical and logical
operations, together with the appropriate and varied
sequences of these, that were to be performed by
equipment capable of carrying on physical (i,e.,
electrical) operations corresponding to the abstract
arithmetical and logical operations . Until the proper
patterns of abstract operations were worked up, no
suitable machinery could be devised.

Everett embodied the abstract operations and
their patterns in "block diagrams" which set forth
the appropriate logical functions. Ideally, once
a block diagram had been organized, presenting the
sequence of logical steps necessary to accomplish,
say, a particular computation, then the engineers
could turn to the problem of designing the electronic
circuits, including the wiring, the resistors, the
condensers, the tubes, and similar elements. These

7.4
circuits when properly constructed could accomplish,
in physical hardware susceptible to differing, con-
trolled, electrical states, the logical steps and
computational results desired.

In essence, the designers' tasks were like
those carried out in the following homely illustration:
to maintain order at a busy street intersection,
colored signal lights are turned on and off in an
appropriate sequence. Any driver who has ever
encountered malfunctioning stop signals knows how
important the orderly sequence is , and any driver who
has waited impatiently for a break in heavy traffic
in order to make a left turn understands how important
it is to install a system of appropriate lights,
appropriately colored and appropriately sequenced to
give the left-turners their legitimate opportunity
to proceed.

The relevance of this illustration to the
designing of computers lies in the fact that lights
turned on and off correspond to the movement of
traffic in different directions, with the result
that a pedestrian at such an intersection, even
when no cars are in sight, knows the meaning of the
pattern and sequence of the colored lights going
on and off. In the case of Whirlwind and contemporary
early computers the sequence and patterning of

7.5
selected radio tubes and circuits turned on or off
meant, or corresponded to, analogous logical and
mathematical operations being carried out. (In
later computers transistors replaced the radio tubes
to carry out the same functions in smaller machines
employing, more efficiently, less electric power.)

The problem that Everett, Forrester, and their
contemporaries faced during the late 1940 's was
that they had little or no experience working
out such sequences ; theirs was the predicament of
auto traffic planners who have had no practical
experience controlling traffic. In lieu of the
knowledge of experience , Everett had at his disposal
the theoretical insights of the pioneering investigators,
among whom were Aiken, Babbage, Bush, Caldwell, Crawford,
Eckert, Goldstine , Mauchly, Stibitz, von Neumann, and
a handful of others. The practical experience of
these pioneers was so limited, in comparison to the
challenge the Aircraft Stability and Control Analyzer
offered, that Everett was compelled to undertake
pioneering and highly complicated system-building of
his own which had no precedent, especially in the
realms of reliability of performance and rapidity
of operation demanded by the simulator.

It is not possible to rank the originality of
Everett's and Forrester's contributions-in-detail

7.6
with those of their contemporaries and predecessors,
other than to point out that Forrester's managerial
and inventive talents and Everett's detailed logical
designs , together with their resulting embodiment
in the assemblage of electronic hardware called
"Whirlwind I , " produced a working computer of
unprecedented speed and reliability and a complement
of engineering personnel possessing unequalled (at
the time) design sophistication and engineering
"know-how." Everett and Forrester, operating as
engineering and managerial alter-egos and supplements
of each other as the years passed, were primarily
responsible for the complexion of the Project and,
consequently, for its failures and successes.

Yet the measure of their contribution to the
state-of-the-art of the emerging scientific technology
of the computer cannot be well assessed for a
variety of reasons , of which the most important is
the lack of balanced assessment of the contributions
of their predecessors and contemporaries after the
brief lapse of a quarter of a century. There has
emerged instead , in the technical computer literature ,
a miscellaneous collection of views which reveals
that the insights of some of the pioneers were
promptly appreciated at the time (e.g., von Neumann's),
others were valued at the time and neglected later
(e.g., Mauchly's), and others were neglected at the

7.7
time and exaggerated later (e.g., the magnificent
failure of Babbage). While these do not exhaust the
range of instances, they illustrate the confused
historical situation existing, a situation that is
a consequence of the prevalence of uninformed notions
regarding the process by which events of the past
give rise to events of the present. The alphabetical
sequence of representative names, given above, corre-
sponds neither to the chronological sequence and the
overlapping of their contributions nor to the relative
value or profundity of importance of their contributions ,
for investigators have not yet carried out the massive
research necessary to clarify the picture and achieve
a consensus. As a result, Aiken's contribution, is
widely hailed, for example, as it should be, while
those of Mauchly or Stibitz or Caldwell -- to cite
other examples -- remain obscured. The problem of
technical and historical evaluation here is basically
epistemological, arising as it does from inadequate
understanding of the historical process, and it is
typical of a technology in which the inventors of
the brick, the wheel, printing, and the telescope,
to name but a few, are lost to history, while the
identity, significance, and roles of the contributors
to the invention of the telegraph, the light bulb,
the radio, television, and many other recent items

7.8
remain obscure because of the relentless and
inappropriate search for heroic figures and the
general oblivion to the social character of the inventive
process in science and especially technology.

For these reasons the most that is attempted in
this history of Project Whirlwind is to lay before
the general reader the managerial, fiscal, and
technical factors that appear to be both distinctive
of the Project and representative of contemporary
research and development. Thoroughly representative
was the procedure by which the Project engineers
proceeded to convert theoretical abstractions to physical
operations carried on by pieces of hardware organized
into a systematic array. Thus, the electronic
circuits had to meet the functional requirements of
the block diagrams . But no computer had yet been
built to specifications such as Forrester and Everett
at MIT and Perry Crawford, Noel Gayler, Harry Goode ,
Leonard Meade, Peter Gratiot, and (later) Captain
O'Rear at SDC contemplated, even though Aiken, Mauchly,
Eckert , von Neumann, Goldstine, and others had
demonstrated both the practical promise and the
theoretical possibility. 1 Furthermore, the operating
speeds required by the ASCA problem were so great as
to be without design precedent. As an early issue
of the Project's Summary Report quietly understated,

7.9
"additional detailed knowledge" was needed regarding
the "timing and synchronization of operations performed
by individual circuits when they are integrated into
large-scale systems. "^ Accordingly, "operating times
for each type of circuit to be used were determined
by measurement , and the block diagrams were redrawn
in terms of these specific circuits. "^

The progress of a single electric pulse through
various component parts of the computer could be
calculated. Consequently, the engineers could
ascertain theoretically whether synchronous operation
of the components was being obtained, modify their
circuit designs to obtain the synchronization they
required, and then test the resulting hardware singly
and in system hook-up to make sure it met their design
requirements .

They found that essential computing operations
could be performed rapidly enough to be acceptable:
"Calculations showed that with present circuits the
multiplication process could be safely performed no
faster than the rate permitted by a time-pulse
repetition frequency of two megacycles per second.
This speed is considered adequate for Whirlwind I,"
Forrester reported at the end of 191+7. Although
faster speeds were attractive and possible, the
engineers realized such modifications would also

7.10
perpetuate design changes and thereby postpone the
operating date . Thus , even though experimental
a-c flip-flop circuits appeared to be appreciably
faster than the d-c flip-flop circuits the engineers
had checked out in detail (fifteen hundredths of a
microsecond, as against twenty hundredths), because
these circuits would be used over and over again as
one of the basic types of building-block throughout
the entire computer and because the engineers knew
too little about the general performance characteristics
of the a-c flip-flops, they would not risk switching,
in premature ignorance, to the a-c design. Besides,
conversion could be accomplished "with little dif-
ficulty if desired" at a later date. 1 *

Realistic engineering policy required continuing
compromises to be made between the attractive,
untried ideal and the practical, in order to achieve
actual machinery. The Project was, after all,
operating on a schedule , a circumstance that neither
the Whirlwind engineers nor the ONR program managers
could ignore, in view of the rising costs of the
Project. The immediate tasks before the Whirlwind
engineers included the formulation of component and
subsystem parameters that would stand, the preparation
of suitable specifications and drawings , and the
delivery of these to Sylvania engineers so that

7.11
physical components and subcomponents could be
manufactured and delivered to the Barta building,
where Whirlwind I would be housed.

At the same time that basic circuit diagrams
were being completed, laboratory testing equipment
was being designed, purchased and developed so that
present and future development of systems could "be
facilitated by a line of standardized electronic
test equipment for generating, gating, and distributing
pulses at desired repetition frequencies . " The
object was to enable research engineers, "by rapid
interconnection of various units, [to] set up and
experiment with sections of computer systems. "^

Professor Murray during his November visit had
raised a question with Forrester that pointed up a
standing problem confronting all computer designers.
Since one defective vacuum tube, one flawed circuit,
could nullify an entire calculational sequence and
possibly an entire program, how would Whirlwind be
protected from tubes or diodes that were about to go
bad, that were becoming marginal in their operation?
While this was not a completely new question to
Forrester, neither was it one to which he had found
an answer until, as he recalled afterward, in the
throes of trying to formulate a reply at that moment

7.12
that would indicate he and his engineers were masters
of the situation, "a solution presented itself. I
realized that by deliberately varying the voltage and
thereby changing the loading on any circuit while
requiring it to carry out a simple operation, a
tube that was losing its capacity to perform would
be forced to reveal its identity under the marginal
conditions imposed. This was how 'marginal checking'
came to be invented. "^

The literature of basic and applied science
is by custom committed to a policy of outlining
for the reader a method of demonstration by which
the asserted correspondence of data to interpretations
and of facts to conclusions may be established. It
is not surprising, in consequence, to find that the
circumstances of discovery and invention usually
vanish unrecorded from history. Thus, the formal
report to ONR indicating that provisions to accomplish
marginal checking would be designed into the machine
contained no reference to the circumstances in which
the technique was invented. Nor was the report couched
in terms particularly calculated to reassure those
skeptical program managers who were aware that they
lacked the "inside" technical view and the visions
of a Perry Crawford, as well as the familiarity with
engineering detail, balanced against a mathematical

7.13

sensitivity that a Murray might be expected to

have as a consequence of his professional experience.

Forrester's official summary report of his

modest innovation was low-keyed, relatively routine

in form, and unexceptionably incorporated as the

last half of a six-paragraph description entitled

"Trouble-location." It is here re-presented in

full from the December monthly report:

Because digital electronic computers
contain many thousands of electronic-circuit
components, failures must be expected. Such
failures almost always cause errors in com-
putation, and temporarily destroy the useful-
ness of the machine. Rapid trouble-location
methods are therefore of great importance .

A scheme which has been proposed for
facilitating the location of faults in WWI
uses prepared groups of test problems whose
answers are known. These problems are of
two types:

(1) Check problems, solved periodically,
designed to use as much of the machine
as possible. Errors in solutions will
indicate that some part of the machine
is not functioning correctly.

(2) Trouble-location problems, designed
to use only small portions of the
machine. Errors in the solution of
one or more of a series of these
problems will provide information on
the location of a fault after its
existence has been demonstrated by an
error in the solution of the check
problem.

The machine itself may thus be made to locate
faults which would require exorbitant time by
manual methods. Simultaneous failure of many
elements , or failure of certain critical elements ,
will result in greater difficulty, but such oc-
currences should be few relative to the total
number of failures.

7. 14

Although primarily intended as a means
for finding steady-state faults due to the
complete failure of a component, this scheme
will be extended to finding of marginal com-
ponents whose complete failure is imminent,
which might be causing random errors . It is
expected that such components can be made to
give steady-state indications of failure by
appropriate variation of circuit supply
voltages and of the repetition frequency of
applied pulses.

As an example, for certain types of faults,
if the voltage of the screen-grid in a marginally
operating vacuum tube is lowered slightly,
complete failure can be produced, permitting
discovery by check problems and subsequent
location by trouble-location problems .

Whirlwind I power-supply systems are there-
fore being designed to permit selective variation
of supply voltages in a range above and below
normal operating values. The added complexity
of cabling and the additional equipment required
for this purpose are believed well justified by
the expected gain in computing reliability. ?

By the following spring the basic requirements

of a marginal-checking system had been worked out,

personnel had been "assigned to design the electrical

and mechanical layouts," and preliminary design

proposals had been composed.^ By the end of that

year (19 48) marginal checking features were being

incorporated in the five-digit multiplier and tested.

If they worked as expected, they would constitute

the basic template, so to speak, of the pattern of

marginal-checking facilities planned for the entire

computer. The five-digit multiplier was the

smallest unit of the arithmetic element that Forrester,

Everett and the others felt they could construct as

7.15
a, representative subcomponent that would early tell
them whether they had a sound building-block of the
vital computational portion of Whirlwind.

It was typical, too, of their philosophy and
mode of engineering procedure : proceed from the
level of system-requirements appraisal to the level
of a consequent component , establish the detailed
design of the latter, build its parts, assemblies, and
subassemblies , testing them singly and together as
they came into being, in order to establish preliminary
operating characteristics , locate deficiencies in
design and materiel, remedy these, and test the devel-
oping component as thoroughly as possible , taking the
time to build whatever special test equipment was
necessary. This procedure, they were convinced, would
obtain a soundly functioning building block of known
operating characteristics that could be depended upon
and rendered compatible to the exigencies of systemic
(as distinct from isolated) operation.

This philosophy of intimate interplay — i.e.,
proceed from logical and mathematical formulations
to design* build, test, integrate, redesign, rebuild,
retest , reintegrate — was a major cause of the
rising costs associated with Whirlwind which strained
relations between ONR and MIT during 1948 and 1949.
It was also part of their engineer's dynamic answer

7.16
to the continuing problem of quality control, and
it was a policy position from which the Project
leaders refused to budge, regardless of the larger
and larger percentage of ONR's contract-research
budget that they kept calling for. From Forrester's
and Everett's point of view, it was the only way to
maintain the high standards the enterprise required
if it was to succeed within a reasonable span of
time. Not only were they convinced it cost less
to do it right than to do it wrong and then engage
in expensive corrections, but they also held a strong
personal commitment to a way of doing based upon
a philosophy of excellence.

At the beginning of 194 8 Forrester had visualized
completion of the computer by stages: the arithmetic
element of the computer would be ready early, and
the five-digit multiplier test portion of this element
would be ready even earlier, for the computational
speeds were likely to take more time than the infor-
mation-in, information-out storage speeds or the
transfer of information from one part to another of
the machine, and it was essential that they fabricate
earlier those parts that would require critical and
perhaps extensive testing. Electrostatic storage
would come last, not because it required little
testing — on the contrary -- but because extensive

7.17
engineering research and development were required.
These would consume the most time, and Forrester
fully realized this.

Time schedules were drawn up for the major
parts of the computer, and from February, 1948
onward, progress toward meeting the schedules
was reported monthly to the Navy. These reports,
comparing actual progress with scheduled progress,
began about the time that ONR embarked on another
intensive analysis of the Project and its operations
in order to establish how it was proceeding. Not
only did von Neumann, presumably at Mina Rees'
request, spend three days during February in the
laboratory, discussing the operations the machine
would be called on to perform, as well as examining
the block diagrams, potential uses, and arrangement
of the projected machine, and in general familiarizing
himself with the state of affairs of the project.
Mina Rees also brought to the Barta building John
Curtiss and H. D. Huskey from the Bureau of Standards,
and they "considered in some detail with Project
Whirlwind Staff the nature of engineering problems
of computer design and the successive stages of
development leading to the final product . "H

By spring Forrester could say that the building
of the computer had begun; Sylvania was fabricating

7.18
components to the specifications of the Whirlwind
engineers. For some units MIT furnished drawings,
for others detailed specifications from which the
Sylvania engineers could lay out drawings and
authorize fabrication. For still others, such as
"the prototype of the 2 8 -tube accumulator panel,"
the Project engineers constructed the first unit to
serve as a model for the Sylvania engineers to
duplicate , and for still others , such as the storage
tubes, the Project maintained its own in-house
enterprise throughout.

By early summer in 1948 tests had revealed that
a standard type 6AG7 vacuum tube lacked the reliability
life span required. Apparently a silicon concentration
in the cathode nickel was raising a barrier to
current flow, so the decision was made to switch to
a tube manufactured under different techniques, the
type 7AD7 , which appeared provisionally satisfactory.
Many tube sockets in the circuits would have to be
changed, but such a prospect was not unusual or
dispiriting in the engineering view of Forrester
and his associates.

Nevertheless, this was one of the factors
that accounted for what had become a five-week lag
behind schedule, over the first six months of 1948.
Forrester announced at the end of that time that

7.19
since the regular semi-annual revision in the time
schedule was at hand, the schedules would be adjusted
and the status of work actually existing in July
would become the new basis . By this technique
Forrester proposed to put his project on a more
realistic schedule and thus compensate , in a pro-
gramming sense, for "procurement delays, necessary
design changes, and heavy demands upon personnel
time..." 12

Although it might appear at first glance that
he was trying to make the Project look good by
engaging in some sort of scheduling legerdemain, he
was in part postponing the completion date and in
part recovering some of the time lost by reassessing
portions of the program and finding ways to "buy"
time by eliminating, shrinking, or clarifying the
details of previously scheduled operations that
required modification in the light of information
acquired. Information gained from the experience
of the preceding year or more placed him in a
position to specify more sharply the delivery
sequence of some items and the physical composition
of other units. Generally speaking, "actual progress had been
made at about three-quarters of the rate as expected in

January. The new schedule extends the work by 30% in

13
recognition of this fact. "

7.20
Forrester recalled in later years that the
detailed manner in which the monthly Summary Reports
kept ONR posted regarding technical problems and
slippage of schedules had made such a virtue of
frankness that one of the unlooked-for effects was
the added fuel they provided to stoke the persisting
unease the ONR programmers felt.1 1 * The same tenor
of events reflected by these reports did not disturb
Nat Sage , Gordon Brown , and the MIT leadership ,
however, although it should not be supposed that the
latter were directly involved and informed as to
details until the increasingly apprehensive protests
from the Navy reached their ears .

During the second half of 1948 the computer
itself began to appear, as racks, subassemblies,
and assemblies of various component and subcomponent
parts of the computer began to be installed in the
Barta building. At the same time, the prospective
complexities of setting up and then achieving full
operation caused Forrester to postpone the final
completion date once again, this time from the end
of 1949 to the end of 1950.

One of the complexities was the stubborn way in
which an efficient, reliable electrostatic storage
tube design continued to elude the researchers'
grasp even while encouraging advances continued to

7.21
be made. Four activities devoted to the storage tube
were included in the schedule charts submitted monthly
to the Navy during the first half of 1948; the
number of such line items jumped to thirteen after
June. Yet this could not be taken as a sure sign
of trouble, for six months later, at the start of
194-9, Forrester had become optimistic that the storage
tubes would be ready sooner than he had earlier
expected, as a consequence of gains made by the
increased emphasis and effort given during recent
months. But earlier, in the summer of 1948, at
the very time Admiral Solberg was applying pressure
on the Institute to proceed with Whirlwind's
development at a more reasonable (i.e., less costly)
rate , it appeared that the storage-tube problem was
bigger than had been suspected, and the slow rate
of progress toward a solution, compared to that
which he had expected, caused Forrester to redouble
his efforts, and consequently, to up the ante again,
to ONR's dismay.

In spite of such vicissitudes, the research,
design, development, fabrication, and testing progress
being made on all fronts caused Project Whirlwind in
mid-1948 to appear, at least to the MIT administrators,
as a healthy engineering project indeed, and one that
the Institute definitely need not apologize for.

7.22
Such judgments, it should be pointed out, did
not and could not derive solely or directly from any
schedule charts of recitation of weekly and monthly
problems encountered and accomplishments effected.
There was by this time so much going on in the Project,
and so many activities were being coordinated in
so many directions , that only an engineer in
Everett's or Forrester's positions could be
expected to possess the authentic and fully informed
"feel" of how Whirlwind was progressing, and it
can always be argued that their stake in the success
of the Project was too great to allow them the
dispassionate acuity of an objective view.

Whirlwind had long since passed the stage at
which its likelihood of success or failure could
readily be judged, and this circumstance became
one of the reasons why the Project was so difficult
for the Navy program managers to either bring to
a stop or place under tighter rein. The most
experienced and sophisticated of administrators and
analysts, possessing top-echelon authority and
influence -- whether a von Neumann or a Solberg or
a Weaver or a Sage or a Compton — were not at
all inclined to take drastic action in either direction.
They knew instinctively that such judgments, either
to give more rein or pull in the harness, were too

7.23
personal, too intuitive, too complex and obscure in
their bases to be communicated easily and convincingly
to another. To justify or condemn Project Whirlwind
on its intrinsic merit, then, was impossible.

From the point of view of basic ONR policy
and the responsibility for enforcement of that policy
that Mina Rees and subsequently Admiral Solberg
shared, the high standards the young MIT engineers
were determined to maintain became standards too
high and too costly to be long endured. It was
inevitable that a difference of opinion should arise
between MIT and the Navy as to whether ONR was
trying simply to bring Project Whirlwind into line
or to kill it off. In this respect the two views
were perhaps irreconcilable.

In any event, as the budget disagreement
sharpened during 1948 and brought MIT's top management
into direct involvement with Forrester's and Mina
Rees' policy dispute, the MIT leaders became aware
they had a partly finished, well-begun computer of
unique design on their hands. They realized also
that since Project Whirlwind could not be judged
properly, by all the parties concerned, on its own
merits alone and since the prospects and perspectives
Sage and Forrester had offered opened up singularly
powerful channels of persuasion, MIT must undertake

7.24
an appropriate educational and thoroughly legitimate
propaganda and informational campaign, reserving
the "muscle" and power of MIT's position and reputation
for any confrontation that might arise.

Accordingly, they requested further detailed
information. President Compton's September request
to Nat Sage for a report on the uses and costs of
digital computers led, as has been said, to a
flurry of activity within Project Whirlwind. Between
September 14 and October 15, four reports — defenses
of the Project although not labelled as such — were
prepared and published. Two versions of the first
report were prepared. One had been prepared and
submitted to President Compton on the 14th; this
was apparently in response to his request of September
8 . The second version was completed and published
on the 17th. It discussed the same material, but
in more detailed and extensive form, providing more
explanatory and illustrative argument and material.
The first report set forth the possible military
applications of digital computers and included an
approximate estimate of the "time and cost to bring
such information systems to useful military
realization." The authors — Jay Forrester, Hugh
R. Boyd, Robert R. Everett, Harris Fahnestock, and
Robert A. Nelson — estimated the time required would

7.25
be about fifteen years and the cost about $2,000,000,000.
The report was not concerned with the digital computer
solely, but rather with the complete system of which
the computer would be but a part, so that the programs
envisioned included "auxiliary equipment, applications
studies, field tests, and training of staff required
to do research and to produce and operate the equip-
ment." The areas of application which could be
foreseen included "air traffic control, integrated
fire control and combat information centers , inter-
ception networks, scientific and engineering research,
guided missile offense and defense, and data processing
in logistics." The total report reflected the most
advanced thinking of the young MIT engineers and the
SDC engineers at Sands Point. -^

Although the first report was primarily concerned
with the application of digital computation systems
to military needs in a general sense , it provided
a defense for Project Whirlwind without referring
to any specific computer development program, if only
by pointing up the advantages to be gained once
successful development of digital computation systems
made possible "the better integration and more
effective use of other military equipment." In this
manner it justified "the diversion of men and

■I c

resources to digital information-system development. IIJ -°

7.26
Forrester and his co-authors were reiterating their
thesis that the potential of the digital computer
was so great and the benefits to be derived from its
use so immense that the costs involved, no matter
how great , were warranted . To them a national computer
development program was as important to the well-being
of the nation as had been those programs which had
led to the development and use of radar and the
harnessing of nuclear energy.

The analysis of the research and development
program essential to the achievement of digital
computing systems reflected the experience Forrester
and his colleagues had gained since they had embarked
upon their own program in 1945. They noted in their
report, perhaps ruefully, that the costs of making
equipment for military application appeared sometimes
to have been "underestimated because of linear
extrapolation of past laboratory programs . " Instead ,
they argued, similar development programs had "grown
exponentially," and they cited the development of
radar as the "nearest parallel."-*-' This observation
although reflecting their own errors as to time
and costs, countered the charges made by the Project's
opponents that it was too expensive in money, time,
and manpower for the benefits it would provide. The
cumulative costs of such programs , the authors

7.27
argued, would be more than repaid by the benefits
the nation would derive from the application of digital
computers to national needs.

The second report , also quite detailed and
extensive, was directly pertinent to Project Whirlwind,
outlining three possible "levels" of operation for the
period 1949 to 1953. This report was completed on
September 21, 1948, one day prior to the conference
with the Chief of Naval Research, and was designed
to serve as a basis of the presentation which the
MIT group intended to make to Admiral Solberg.
Plan 1 indicated the extent of "the research,
development and limited experimental computer operation"
possible under the proposed budget of 1.8 million
dollars per year . Plan 2 was based upon an annual
operating budget of 3.8 million which would allow
the addition of "a substantial operating force for
the efficient solution of engineering and scientific
problems . " Plan 3 which proposed an annual budget
of 5.8 million would further permit the inclusion of
a research program into "the application of digital
computers to the field of control and military uses."-'-"
To some extent, the second report complemented the
first by noting how Project Whirlwind's program could
be organized to permit the realization of the military
applications outlined in the report of September 17.

7.28
It is interesting to note that the second report
contained no discussion of any program which could
be conducted under the minimal figure of 1 . 8 million
dollars. The minimal rate considered was that proposed
by Sage in his letter of August 4 concerning the
allocation for Fiscal Year 1949 which had been based
upon an anticipated average monthly expenditure of
$150,000. There seems to be no doubt but that the
directors of Project Whirlwind were determined not
to strike their flag, if strike it they must, without
a battle. Forrester recalled in after years that
the Whirlwind group had by this time become so
deeply committed to the idea of doing the technical
job right or not at all ("Do it on our terms, or let
it be shut off!"), that the arbitrary, unilateral
nature of the view they took was not readily apparent
to them. Instead, they were aware that there were
other worthwhile projects to which they might apply
their talents, should the Navy find itself unable
to supply the proper funds. -^ Their high spirited
mixture of determination and bravado was not put to
the test on this occasion, however.

September 22, 1948 had been agreed upon as the
day for the Navy and the MIT representatives to meet
in Washington to discuss the financial and technical
ramifications of Project Whirlwind. ONR was sufficiently

7.29
impressed with the importance of the conference to
hold a "rehearsal conference" on the 21st, a rehearsal
that lasted all day but which, when compared to MIT's
preparation, was barely minimal. The purpose was to
acquaint the Chief of Naval Research with the program,
but more importantly, perhaps, "to establish a common
understanding within the Office of Naval Research and
to solidify the thinking of all individuals within
the Office of Naval Research on the Navy's position
relative to Project Whirlwind." 20 Thus, ONR was
establishing its "party line" even as MIT, through
the conference called by President Compton on
September 8, had established its "party line."
Each organization had taken a tentative position for
the first round of discussions, but each, as events
were to prove, had also remained sufficiently flexible
to permit compromise .

The general conference of the 22nd served in
many ways as a forum for the reiteration of previous
questions and explanations. Forrester explained the
reasons underlying the transition of emphasis from
an aircraft simulator to the digital computer,
covering the same ground he had covered many times
before. Mina Rees once again related the questions
of comparative costs between Whirlwind and the
computer von Neumann was developing at the Institute

7.30
for Advanced Studies. Captain J. R. Ruhsenberger ,
Director of SDC, "presented an emotional plea for
the aircraft analyzer and the benefits that would
accrue from it." Perry Crawford, to Forrester's
dismay and irritation, was strangely quiet about the
many conversations he and others from SDC had held
with Forrester and his associates concerning the course
Project Whirlwind was following and the varied and
diverse applications open to it, leaving the "erroneous
impression" Forrester later noted, that SDC "had been
steadily interested in the aircraft analyzer problem
to the exclusion of other applications. "21

What Forrester perhaps did not know or recognize
at the time was how greatly the visionary spirit with
which SDC had infected Project Whirlwind had waned
within the Navy. SDC had lost its fight against ONR,
and Crawford's silence was in part the silence of
dejection and in part the continuation of a policy
that de Flore z had adopted at the start. For al-
though de Florez's experience with aircraft simulators
and the insights he had picked up from Hunsaker, whom
he had known since college fraternity days , all led
him to place greater emphasis in his own mind upon
the design aid that a simulator run by a computer
could render and less emphasis on its undoubted
virtues as a training aid, he had nevertheless always

7.31
led from strength when appealing for support from
within the Navy by stressing the trainer application.
Crawford in later years reflected that probably by
early 1948 de Florez had given up seeking a broader
mission for SDC. 22 The aggressive, visionary spirit
infecting Special Devices personnel in 1947 and
responsible for encouraging Forrester and Everett,
in frequent meetings at Sands Point, to see more
ambitious prospects of the sort that had stimulated
the forward-looking systems-control views represented
by their L-l and L-2 Reports, had all but disappeared
from SDC by the time of the ONR — MIT meeting in
September of 1948.

The important point made by ONR in the course
of the September discussions was its inability to
meet the financial requirements of the Project as
set forth by its directors. Indeed, the question was
raised if the Project as envisioned by the Institute
was not too large for ONR to handle and, possibly,
too large even for the Navy. The ceiling of $900,000
already established was to the maximum for ONR, and
the MIT representatives were asked to determine "how
the program could be continued for the fiscal 1949
period for that amount." The reply, made by Nat Sage,
was that "no immediate or tentative solution was
foreseeable." Nevertheless, the expenditure rate of

7.32
$150,000 per month would be reviewed in the hope a
reduction could be effected. Then Sage left the door
open by volunteering the observation that an allocation
of $1,200,000 rather than $900,000 "would probably
be sufficient to finance Project Whirlwind to the
end of Fiscal 1949 (30 June 1949 )." 23

The larger amount, when added to the $385,260
carried over from Fiscal Year 1948, would permit
an average monthly expenditure of approximately
$132,000, some $18,000 under the anticipated rate of
$150,000. It was an amount which would undoubtedly
permit the program to continue without any drastic
cutbacks, although the rate of acceleration would be
less than Forrester could have preferred. If the
additional monies were allowed by ONR -- and eventually
they were — then the total funds available to
Project Whirlwind for Fiscal Year 1949 would be
approximately a quarter of a million less than the
original request.

Despite the adamant stand the ONR representatives
took regarding the $900,000 ceiling, the conference
concluded with the formulation of an agreement signed
by the representatives of ONR and MIT, which strongly
implied that additional monies would be forthcoming,
provided MIT would strive to hold costs to the minimum
by a "reasonable diminution of effort." The $900,000

7.33
would be formally allocated. The Project's estimated
date of completion would be extended to April 1,
19H9; however, every effort would be made by Forrester
to stretch the allocation to cover as much as possible
of the three months between April 1 and the end of
the Fiscal Year, June 30, 1949.

MIT had been granted the $900,000 for a nine
month period upon the understanding frugality would
be practised. Meanwhile, evaluation of the Project
would be continued, to determine what would be the
"best reasonable rate of effort on a scaled-down
basis for future operation." 2*+ A week later to the
day, Amendment No. 6 to Task Order 1 of Contract
N5ori-60 was issued, extending the date of completion
and confirming the allocation of $900,000.

The conference ended in a compromise, leaving
for future discussion and solution the final resolution
of the rate at which the Project would be conducted.
MIT had received assurances of continued support,
even if not to the extent desired. ONR had received
assurances that measures would be taken to limit the
Project's rate of acceleration and had succeeded in
reducing its allocation without seriously offending
the Institute. This latter was without doubt a very
serious and sensitive consideration for ONR as it
sought to establish and retain the growing confidence

7.34
of the academic community in ONR's ability to mount
sustained, consistently managed and funded research
programs .

The agreement which ended the conference was
temporary and expedient, pending final evaluation and
decision. Subsequently, Solberg emphasized this view
in a communication to Compton, expressing his
conclusion that the Project was a "long-range one,"
to be evaluated within the context of the total
national computer development effort as well as
within the context of ONR's total research program.
Solberg was striving to be fair. He was willing
to accept temporary continuation of the program at
a rate which approached ten percent of ONR's University
Research Program funds and even to consider the
allocation of more funds if absolutely necessary.
He was not willing, however, to grant unlimited funds
and freedom of direction, at least not until a
thorough investigation of the Project had provided
a sound appraisal of Whirlwind ' s genuine importance
and position within the total national computer effort. 25

Compton in an "off-the-record" reply to Solberg
explained that future discussions on Project Whirlwind
would be conducted for the Institute by James R.
Killian who, upon Compton 's resignation to succeed
Vannevar Bush as Chairman of the Research and Development

7.35
Board, would become president of MIT. Observing
that in light of his new appointment , he could not
properly be "an intermediary in these discussions,"
Compton nevertheless did informally convey to the
Chief of Naval Research some thoughts which he believed
Killian would later express concerning the Institute's
stand in the matter. In general he concurred with
Solberg's opinion that the government computer
development program was so important and costly that
it deserved "the most expert possible evaluation,"
pledging the Institute in the meantime to respect
the agreement reached at the Washington conference.

Turning to Project Whirlwind, Compton explained
that "our group" -- the term was his — was preparing
a memorandum for the Chief of Naval Research which
would "add considerable clarification of the issues"
for Solberg as it had for Compton. The memorandum
would explain the "philosophy" of approach taken by
Project Whirlwind. As Compton saw it, this approach
appeared to differ in three respects from other computer
development programs, "especially the one at Princeton."
He was convinced , he wrote Solberg , that the IAS and
MIT programs were "essentially non-competitive in
the sense that one may prove to be a useful research
tool and the other a useful operational tool."
Through his informal reply, Compton permitted Solberg

7.36
to infer not only the Institute's position, but also
and perhaps more importantly, the position which the
Chairman-designate of the Research and Development
Board would probably take . In a sense , Solberg was
being forewarned by Compton.^°

NOTES TO CHAPTER 7.

1. Interview, P. 0. Crawford by the authors, October
25, 1967.

2 ' Project Whirlwind ( Device 24-X-3 ) , Summary Report
No. 3_, Dec. 1947, p. 2. Hereafter cited: Summary
R"eport No. -. These reports began to be submitted
on a monthly basis, beginning in Nov., 1947, to
"the Special Devices Center, Office of Naval Re-
search, under Contract N5ori60" by the Servo-
mechanisms Laboratory.

3. Ibid . , p. 3.

4. Ibid . , p. 5.

5. Ibid .

6. Interview, J. W. Forrester by the authors, July
24, 1964.

7. Summary Report No . 3^, Dec. 1947, p. 8.

8. Summary Report No . 6_, Mar. 1948, p. 19.
9 * Summary Report No . 15 , Dec. 1948, p. 19.

10. Summary Report No . 3_> Dec. 1947, p. 7.

11. Summary Report No . 5_, Feb. 1948, p. 3.

12. Summary Report No . 9_, June, 1948, p. 8.

13. Summary Report No . 10 , July 1948, p. 2.

14. Interview, J. W. Forrester by the authors, July
31, 1963.

15. J. W. Forrester, H. R. Boyd, R. R. Everett, H.
Fahnestock, and R. A. Nelson, "Forecast for Military
Systems using Electronic Digital Computers,"
Report L-3, Servomechanisms Laboratory, MIT, Sep-
tember 17, 1948; the first version prepared by

the same authors was entitled, "A Plan for Digital
Information Handling Equipment in the Military
Establishment," Report L-3, Servomechanism
Laboratory, MIT, September 14, 1948.

16. Report L-3 , September 17, 1948.

17. Ibid.

18. J. Forrester, H. R. Boyd, R. R. Everett, H.
Fahnestock, and R. A. Nelson, "Alternative Project
Whirlwind Proposals," Report L-4, Servomechanisms
Laboratory, MIT, September 21, 1948.

19. Interview, J. W. Forrester by the authors, October
26, 1967.

20. Memorandum from J. B. Thaler to Director, Con-
tract Division, SDC, Sub j . : "Trip to Office of
Naval Research, Washington, D. C. on 20 September
1948 concerning Contract N5ori-60; Report on,"
September 24, 1948.

21. Ibid.; J. W. Forrester, Computation Book No. 45 ,
pp. 134-5.

22. Interview, P. 0. Crawford by the authors, October
25, 1967.

23. J. W. Forrester, Compu tation Book No. 45 , pp.
134-5.

24. Ibid . ; Ltr. , T. A. Solberg to K. T. Compton,
September 29, 1948.

25. Ltr., T. A. Solberg to K. T. Compton, September
29, 1948.

26. Ltr., K. T. Compton to T. A. Solberg, October
7, 1948.

Chapter Eight
R&D POLICIES AND PRACTICES

The memorandum to which President Compton referred
in his letter to Admiral Solberg emerged in definitive
form as two reports -- the third and fourth in the series
of four generated by Compton' s desire that the nature and
purpose of Project Whirlwind be clearly articulated. Bear-
ing the dateline October 11, 1948, both reports sought to
emphasize the importance of Project Whirlwind by explain-
ing the unique characteristics of the Project and the con-
tribution it had to make to contemporary computer technology.
The third report, "Memorandum L-5," set forth the general
philosophy and plan of attack which the Project sought to
follow. The fourth report, "Memorandum L-6," offered a
comparison between MIT's Project Whirlwind and the computer
program at the Institute for Advanced Study in Princeton,
New Jersey .

The third report, which was prepared by Jay Forrester
himself, set the Project within the context of the policies
and procedures of the Servomechanisms Laboratory in an
effort to show how the program that Project Whirlwind was
following reflected the purposes and procedures of the
Laboratory and, by implication, the very principles which

8.01

8.02
MIT itself followed. Noting the preference of the Laboratory
for projects which combined "engineering research and devel-
opment with systems consideration," Forrester sought to point
up the design, development, and construction of Whirlwind as
but one element of a system which was, in this instance, by
contractual agreement an aircraft analyzer. At the same
time, he reiterated a favorite argument: "the scope of the
project might not be justified by this application alone,
were it not for the benefits which will accrue to all other
digital computer applications." These were applications which
Forrester felt could be so important that the Navy might even
decide to "redirect future work." To Forrester this meant
the natural development of sophisticated "control systems"
for practical military use in the future.

Time was of the essence, he argued. It could not be
wasted by following the usual sequential procedures of
"research, development, and design." These three steps had
to overlap, even run concurrently when possible, in order to
obtain a "reliable operating" computer at the earliest pos-
sible moment. Herein lay the singular strength (and cost-
liness, it should be admitted) of Project Whirlwind, for
even as the Project conducted research, it built and tested.
In addition, through its use of graduate students in the
tradition of the Servomechanisms Laboratory, it produced

trained and experienced personnel for "a development of

3
national value."

8.03
Forrester and Everett jointly prepared the fourth report
as a rebuttal to the charges of those who had persistently
implied that the digital computer project at MIT was inferior
to the von Neumann project at the Institute for Advanced
Study. The two authors argued that the two programs had been
established for different purposes and consequently followed
different procedures. About the only thing they possessed
in common was the intent to design and construct a "parallel-
type digital computer;" otherwise, the two groups held "very
few common views on the methods for specifically achieving
working equipment."

As von Neumann wisely had done before them, in his re-
marks to Mina Rees regarding Professor Murray's report,
Forrester and Everett made no attempt to demonstrate the
superiority of their program by denigrating the IAS program.
Rather, they recognized the differences between the projects
to be quite valid, for their origins lay in different pur-
poses and projected uses. Von Neumann and his associates at
the Institute for Advanced Study were "engaged in scientific
research . . . the study of high-speed computing techniques;"
Forrester, Everett, and their associates at MIT were "engaged
in engineering development ... to produce and use computers."
Moreover, von Neumann was seeking to design and construct a
digital computer; the young MIT engineers were seeking to
design and construct a system employing a digital computer
as an integral element.

8.04
Speed and reliability were of greater importance to the
MIT program, since the digital computer within the system
had to operate "in real time" and with minimal error. The
IAS program, on the other hand, since it was to be used pri-
marily for mathematical computation did not need to meet the
same standards of speed and reliability. These differences
in purposes and goals made necessary a difference in pro-
cedures that in turn led to a difference in costs. Project
Whirlwind was building a "prototype," and although the
approach followed was "less efficient and more expensive,"
it was faster, and this was a consideration of primary im-
portance under contemporary conditions.

While Solberg and Compton were exchanging views and
Forrester and his associates were preparing reports, the
press for additional funds from ONR continued. The $900,000
which ONR had allocated for the fiscal year 1949 was some
$300,000 short of the amount Nat Sage had proposed during
the fall as an acceptable compromise. Forrester had accepted,
albeit reluctantly, a $1,200,000 ceiling and with his staff
had planned a program conforming to an allocation in that
amount. He aggressively sought to prevent any further re-
duction in what he already considered an inadequate budget.

In addition to forestalling further cuts, he had to
convince ONR that additional funds were mandatory if a mini-
mal rate of progress was to be maintained. To this end,
Forrester and members of his staff prepared in massive detail

8.05
position papers which explained, on the one hand, the financial

needs of the Project for the balance of fiscal year 1949 and,

on the other hand, the disastrous impact upon program goals

which would follow if the Navy failed to meet those needs. 5

In addition to preparing position papers that would con-
vince both MIT's top management and official government
circles of the immediate and vital importance of computer
development programs in general and of Project Whirlwind in
particular, Forrester became more active and more personally
involved in "selling" his Project to his more influential
and recognized colleagues within the Institute. Early in
November he spent a morning with Dean T. K. Sherwood of
Engineering, Nat Sage of the Division of Industrial Cooperation,
and Professors Harold L. Hazen, Jerome B. Wiesner, Samuel H.
Caldwell, and Gordon S. Brown, "discussing applications of
Whirlwind I, particularly to scientific problems and to con-
trol applications."

With the exception of Caldwell, those present at this
meeting were to participate in a subsequent conference in December

with representatives from ONR. Meetings of this kind not only

reflected Forrester's desire to win the support of the more
influential members of MIT's academic community, but also
Comp ton's intent to have the Institute's top scholars in areas
directly pertinent to the work of Project Whirlwind become more
familiar with its nature and purpose and with its director.

8.06
Three days before the meeting with ONR, Forrester had
lunch with Professor Hazen and used the occasion as an
opportunity to explain that the "lack of mutual understanding"
present between ONR and Project Whirlwind found its origin in
the different approaches each took to computer research and
development, approaches based upon their respective views
concerning the ultimate use of the computer. Mina Rees and
her associates approached the matter from the view of the
mathematician, whereas Forrester and his associates approached
it from the view of the engineer. Since Hazen was planning
to have lunch with Mina Rees and E. R. Piore, Deputy for
Natural Sciences, ONR, Forrester apparently hoped Hazen would
become an intermediary and try to make clear the validity of
the engineering approach and possibly clear up some of the
misunderstanding. Other causes for the misunderstanding were
probably discussed also, for Forrester entered in his record
of the lunch, the cryptic comment that the two men has also
"covered . . . the current political situation."

The persistent pressure for more funds, Solberg's de-
sire to gain a greater insight into the nature and purpose
of Project Whirlwind, and the mutual intent to resolve the
differences between ONR and MIT stemming from the conduct of
the Project led to another conference between representatives
of the two organizations at Cambridge on December 9, 194-8.
Both organizations sent their top staff to the meeting.
ONR was represented by the Chief of Naval Research, Rear

8.07
Admiral T. A. Solberg, accompanied by Dr. Alan T. Waterman,
Deputy Chief of Naval Research and Chief Scientist; Dr. T.
J. Killian, Science Director; Dr. Mina Rees, Head of the
Mathematics Branch; and several others, among whom was Perry
Crawford, then on temporary duty with the Research and Develop-
ment Board.

MIT was represented by its President-designate, Dr.
James R. Killian, Jr., accompanied by Dr. T. K. Sherwood,
Dean of Engineering; Nat Sage, Director of the Division of
Industrial Cooperation; Professor Gordon Brown, Director of
the Servomechanisms Laboratory; Professor H. L. Hazen, Head
of the Department of Electrical Engineering; Professor J. B.
Wiesner, Assistant Director of the Research Laboratory of
Electronics; and Forrester and other members of Project
Whirlwind's staff. The official positions and quality of the
representatives alone gave proof of the importance both ONR
and MIT attached to the meeting and to the matter under dis-
cussion.

The meeting was chaired by Dean Sherwood, who also acted
as the chief spokesman for the Institute. After a few re-
marks by Forrester on the program and its rate of progress,
the group from Washington toured the laboratory . Then the
two groups settled down to a serious discussion of the matters
at issue. The exchange of views was quite blunt. Sherwood,
in order to counter rumors to the contrary, "stressed the
united MIT support of the Project," and noted the "desirability

8.08
of having good communication among all groups concerned to
prevent the spread of rumor, and especially of the need of
having some technically competent person in Washington desig-
nated to follow the Project." The very presence of MIT's
scholars and administrators gave substance to Sherwood's
point.

The conferees acknowledged that "confusion" had been
created and "some appreciation of background lost" with the
transfer of supervisory authority from SDC to the Mathematics
Branch of ONR. Mina Rees acknowledged that her ignorance in
engineering made her incapable of comparing the Whirlwind
program and its emphasis upon an engineering approach with
other computer projects which sought different goals and
followed different procedures. Consequently, she expressed
her intention to have the Project "evaluated by independent
experts . "

When the discussion turned to financial requirements,
Nat Sage estimated the additional funds necessary from fiscal
year 19M-9 funds to approximate $275,000. Sherwood, in his
comments on funding, left the ONR representatives with a
thinly veiled warning by explaining that it was MIT's policy
to give three months notice when releasing staff members.
Since there remained only sufficient funds to continue op-
erations for four months, he suggested a "prompt decision"
concerning the allocation of additional monies be forthcoming
from ONR. Bending to this pressure, the ONR group intimated

8.09
that more funds would be made available and requested a state-
ment on the amount necessary to get the Project through fiscal
year 1949. In light of the blunt exchange of views and ONR's
acceptance of the need for more funds, it is small wonder
that Forrester expressed the conclusion that the "general
result of the meeting seemed to be quite satisfactory."^
Perhaps in some ways he was a bit optimistic, for although
the December conference brought out the full strength of MIT,
it also marked the apogee of the support which the Institute's
directors were to render Project Whirlwind.

The following day Nat Sage submitted to ONR a request
for additional funds in the amount of $378,186 to carry the
Project to June 30, 1949. The Navy ultimately made available
$300, 000, 8 providing thereby a total of $1,200,000, the
amount which Sage upon several occasions had suggested would
be acceptable. In a letter informally notifying Sage of the
additional funds and the extension of the contract to June
30, 1949, Mina Rees added that budgetary considerations in-
dicated ONR would be unable to allocate to Project Whirlwind
more than $750,000 for fiscal year 1950. Consequently, she
advised, it was essential the Project "eliminate . . . any
long range activity, supported by ONR, which does not contribute
in a direct way to the completion of Whrilwind I." That this
could be done, she explained, had already been determined in
conversations with Forrester in Washington in early January, 1949.

8.10
In his conversations with Mina Rees, Forrester had
acknowledged that research could be terminated if necessary
to "conserve funds for the completion" of the computer, but
his point of practical reference was the continuation of the
$100,000 per month rate of expenditure which he and his col-
leagues had accepted as the minimum amount for the maintenance
of a dynamic and active program. This was the amount required
to underwrite the anticipated program outlined in the draft
of Memorandum L-10, which Forrester and Mina Rees had before
them. He predicted that if this rate were continued, the
computer could be completed, using a test storage component
by October 1, 194-9 and an electrostatic storage component by
December or January. Mina Rees asked if it were not pos-
sible to continue this rate of expenditure for six or eight
months and then taper off; this Forrester acknowledged as a
possibility. Mina Rees and C. V. L. Smith of ONR, who joined
the conversations, were both of the opinion that $1,000,000
was the maximum which could be anticipated for fiscal year
1950. They recommended that the program be planned on that
basis. Forrester was left with the impression that the Chief
of Naval Research was unwilling to approve a larger amount;
for that matter, the figure mentioned did not yet have his
approval.

It seems clear, from the discussion between Forrester
and Mina Rees and the subsequent allocation of $750,000 for
Project Whirlwind, that the directors of ONR had determined

8.11
to reduce the costs of the Project to a level that would place
the allocations more in line with the monies available to ONR
for research and development in computers. Whether this was
a general goal or Solberg's alone cannot be determined, but
certainly Forrester's comments following his discussion with
Mina Rees convey the impression that the Chief of Naval
Research was exercising a strong influence in this direction.

Apparently, Solberg had become convinced that the time
had arrived to terminate research and push the Project to
completion at minimal cost. Political as well as technical
considerations made necessary a more gradual reduction than
perhaps he preferred, but the evidence suggests he had no
intention of underwriting the Project at the level proposed
by Forrester. Subsequent actions by the leaders of MIT
suggest that, fully recognizing the problems besetting ONR,
they were more amenable to a compromise than was Forrester,
with the result that he found himself being placed under
greater pressure within the Institute itself. The creation
of a computer center at MIT appeared to be an instance of
this pressure, but it was an operational development that
Forrester himself suggested in order to facilitate the ef-
ficient operation of the Whirlwind project.

Neither position papers nor oral argument were success-
ful in moving the Navy, however. ONR allocated to Project
Whirlwind for fiscal year 1950 not the one million dollars
that Mina Rees and Forrester had talked about early in 1949,

8.12
but three-quarters of a million, and Forrester had to tailor
his program accordingly. Fortunately, his luck, or that of
the Project, was still running strong. Since July of 1949,
conversations had been taking place between Forrester and
his associates and representatives of the Air Force over the
possibility of applying the digital computer they were develop-
ing at MIT to air traffic control. These conversations led
eventually to the negotiation of a contract in the amount of
$122,400 for research in this area during the period from
March 1, 1949 to April 30, 1950. 10 Although the amount of
money involved was not relatively large, it did help amelio-
rate the situation, if only by providing a means whereby key
professionals could be kept under salary. More importantly,
however, it was to lead to the creation of a pool of experi-
ence which proved of immeasurable value when the Air Force
later turned to MIT and the rest of the national educational
establishment in its desperate search for air defense techniques
and equipment.

With the funds available to- him for fiscal years 1949
and 1950 Forrester was able to push the Project steadily
on toward completion. Most of the people on the Project
became aware of the pressure that the Mathematics Branch of
ONR was applying, but few outside of Forrester, Everett, Boyd,
Fahnestock, and others who had been there since 1946 realized
how far the Navy had shifted from its once enthusiastic
support of ASCA to its grudging funding of Whirlwind. To the

8.13
old hands, the crisis of the fall of 1948, the recurring in-
spection-visits by different outside experts, and the attitude
of ONR and its Boston Branch Office personnel made it clear
that the old happy days were gone. The Project was being
put on its mettle.

There was so much to be done, inside the laboratory, the
problems were so new and so challenging, the materials needed
were sufficiently available, and the signs of design progress
were so encouraging that philosophy, attitudes, and morale
remained optimistic, confident and constructive, for the most
part, among the working personnel. In one respect, the
engineers, the graduate assistants, and the technicians, were
surrounded by nothing but technical problems, but they saw
these, variously, as easy to solve, or fascinatingly chal-
lenging and ultimately soluble, or as sufficiently stubborn
and unyielding to require alternative appraisal and a shifting
of the line of attack in order to alter the problem to soluble
form.

Forrester was sufficiently optimistic about the progress
they were making to draft a memorandum to Sage in January of
1949, declaring that the research phase of the project was
virtually complete and that "only a small amount of design,
a fair amount of construction, and installation remain to be
finished." The subassemblies of the computational heart
of the computer, the arithmetic element, had been completed
and tested separately and were being linked together to form

8.14

12
the element itself. Marginal checking techniques were being

tested on circuits of the five-digit multiplier and yielding
promising preliminary results. While only three of the four
electrostatic storage tubes fabricated in December functioned
well enough to submit to circuit and life tests, 13 such in-
cidents in engineering development operations were to be
expected, especially since the fabrication techniques for
these special, giant cathode-ray tubes were as experimental as
the very tube designs themselves.

Multiplication and shifting operations were attempted
first in the arithmetic element, and as summer approached,
further testing indicated that the speed, versatility, and

reliability sought in the design phase were being approached

14-
in preliminary operational phases. In the meantime, type

7AD7 vacuum tubes were found to be deteriorating sufficiently
rapidly (in the five-digit multiplier, for example) to warrent
investigation of the causes by Sylvania and MIT engineers.
Since the 7AD7 pentodes, together with 7AK7 gate tubes, com-
prised about two- thirds of the 4-, 000 tubes Whirlwind I was
expected to require, project engineers began to ride the
problem closely. They had already shifted from type 6AG7
tubes earlier, In an attempt to eliminate this problem.
Impurities in the coating materials applied to the cathodes
were again the target of their studies, and circumstances of
the fabrication of various batches of these tubes by the
manufacturers came under close scrutiny. - 1

8.15
By the end of 1949 they were able to point to certain
alloys used in tube manufacture that continued to support the
theory that silicon was a cause of deterioration, but the
sources of the silicon traces were not always easy to determine,
nor was there evidence enough yet to demonstrate how tubes
should be fashioned to guarantee a tube life running into the
thousands of hours desired if tube replacements were to be
kept from occurring at too high and impractical a rate. Thus,
although knowledge of the causes of tube failure seemed to
be sufficient, engineering fabrication and the "reduction to
practice" of that knowledge provided challenging problems to
which only continuing analysis and controlled life-testing
would provide acceptable, long-run, practical answers. Such
problems, with their preliminary solutions and long-term
resolutions, were characteristic of the tenor of technical
events in the Project during 1948, 1949, and 1950, and they
remained unaffected by the funding adventures generated by
changes in national (as well as Navy) fiscal and program
policies.

The momentum that Project Whirlwind had generated by
1949 largely determined the character of its operations during
the remainder of the time that the Navy remained the principal,
federal program manager and during the first years after the
Air Force stepped into the picture. It was not until the
mid-1950 T s that the larger momentum of continental-defense
policy needs, which caused MIT to create Project Lincoln and

8.16
the Lincoln Laboratory, presented a superior force. To this
force the Project stubbornly bent, then yielded, modified its
operational character in the transformation, and lost first
control over its own destiny and then its original leader.

But in 19"49 the individuality and the dynamic character
of the Project as a group of young men organized to carry on
specialized electronics research and development were in full
strength, and the sources of this vitality were to be found
not only in the personal qualities of the leadership Forrester.
Everett, and their group leaders provided, essential though
these were, but also in the technical procedures and policies
that Project personnel followed in carrying out the technical
tasks and resolving the technical problems that arose.

It would be presumptuous to state which conditions,
which procedures were essential and which were not, but it
is possible to describe the conditions that prevailed. From
among these, as well as others not mentioned, one might
imagine, select, and combine the elements of a research and
development enterprise of commanding efficiency and excellence
of performance. Further, one might operate it under enlight-
ened philosophies of costs, of resources, of management, and
of goals. In these respects, the virtues as well as the de-
ficiencies of Project Whirlwind and its mode of operation may
prove useful to examine here.

As has been indicated, there was a persistent search for
talented and intelligent personnel. Continuing efforts to

8.17
keep standards high were matched by continuing efforts to
supply whatever materials were needed. Quality personnel
required quality supplies if they were to be kept busy and
if the policy were not to degenerate into a mere slogan. The
Project was allowed no choice but to follow the wills of its
masters and operate in an atmosphere calculatedly kept as free
as was humanly possible from the exasperations arising from
delays caused by lack of suitable materials or by inadequate
performance of inferior substitutes. Preliminary design and
testing obviously could not avoid every exigency that might
unexpectedly appear, and some unplanned-for delays were to be
expected while an elaborate piece of test equipment, for example,
was being ordered or built to cope with a new technical re-
quirement. Shortages of standard supplies, however, were
considered inexcusable because of the very fact that they were
avoidable. The man who lacked the care or the ability or the
pride to avoid the avoidable found the Laboratory too stren-
uous and soon left, either by choice or invitation. The
Project placed a premium on foresight and careful anticipation
of needs. To encourage these, it provided a more expensive
working climate of planned yet prudent plenty, in which
efficiency, morale, and productivity prospered, than (one
could argue) would have been necessary in a "business-as-
usual" operation.

Along with this philosophy of plenty ("hot and cold
running secretaries," as one critic sarcastically put it),

8.18
went a philosophy of prudence and accountability. Office
walls were to be kept clean and bare of cartoons and frivo-
lous pictures. Verbal and written reports in quantity were
insisted upon, and since the immediate availability of these
was considered imperative because of the need to circulate
the technical information they contained, the Project soon
had its own print shop and photo lab. Friday afternoon work-
bench clean-up resulted in incidental inventory and thus kept
work space from degenerating into storage space. One never
knew when Everett or Forrester would stop by a work bench or
a test rig to see what was going on; there was no question
they were keeping in touch, nor was there any reason to doubt
their ability to grasp the essentials of a problem and see
promising avenues of attack.

"Their approaches were very different," reminisced one
engineer. "Bob Everett was relaxed, friendly, understanding--
and I have never seen anyone who could go right to the heart
of a problem so fast! Jay was as fast, maybe faster, but
he was always more formal, more remote somehow, and you
weren't always sure how dumb he thought you were, or how
smart. That kept us on our toes, I suppose. It was difficult
to know what he was going to do next, but he was so terribly
capable, it didn't matter if you couldn't follow his reason-
ing. He was always thinking with seven-league -boots on. It
made him a pretty formidable guy to work for--partly because
he and Bob always made sure you understood the problem you

8.19
were working on, by finding out what you didn ' t know as well
as what you did know, if you get what I mean. I never re-
sented Jay's obvious ability, but he wasn't the sort I'd call
easy to work for. He definitely never was 'one of the boys.'
He was the Chief, cool, distant, and personally remote in a
way that kept him in control without ever diminishing our
loyalty and pride in the Project, somehow."

"Forrester would come into the lab and tear everything
apart," recalled another with a smile, "and Bob would come
along and put it back together again."

"Tear you apart, you mean," said a third.

"Well, maybe so.... There was absolutely nothing per-
sonal about it, though. He was not an easy guy to know. No
small talk, or if there was, it was such an obvious preamble
to getting down to business! The chances were, your problem
was one he'd run into before somewhere and found the answer
to, and I never could see how he could be so patient. There
was no question who was boss. You took it for granted he
could design anything you could faster and better, but then,
I was a graduate student privileged to work in a hush-hush
classified project--that was before Korea changed everything —
and to my eyes at that time he had had an awfully impressive
amount of experience, from World War II days on. Bob had,
too, but somehow I was never in awe of him the way I was
with Jay."

8.20

"There were Jay's Friday afternoon teas," recalled the
third. "I remember feeling I'd really arrived when I was
asked to attend. That was later on, and he kept the group
manageably small, so that whoever was reporting was talking
about something of use or relevance to your work. Like every-
thing Jay did, it was run very efficiently — very high signal-
to-noise ratio!"

Nearly a third of the technical workers (as distinct
from supporting and clerical personnel) were graduate students,
seeking or working on thesis topics. Their participation
produced highly motivated, rather than perfunctory activity,
and as they were brought up to a sufficient level of familiar-
ity and competence to handle the short-wave, video-pulse
phenomena around which the brand-new type of machine (Whirlwind
I) was being built, they were set to work on particular
problems in solo fashion.

The lines of investigation were many, relative to the
number of staff, and Forrester and Everett keenly realized
that the cessation or interruption of any individual's work
could bring to an abrupt end one of many concurrent courses
of inquiry that were essential to the continuing progress
of the project. Moreover, if the investigator were to leave,
because of a cutback in funds, his work could not readily be
picked up by another because it was not routine.

Since the fundamental business of the Project was probing
the engineering unknown, research to obtain engineering

8.21
specifications and parameters was regarded as an essential
preliminary to design, and wherever possible, design was to
incorporate advance provisions for testing. It was an ide-
alized principle in Professor Gordon Brown's Servomechanisms
Laboratory that the gamut should range from research and
creative design through a practical, working prototype, and
Project Whirlwind held tenaciously to this principle;
"experimental equipment, merely for demonstration of principle
and without the inherent possibility of transformation to
designs of value to others, does not meet the principle of
systems engineering."

Systems engineering required the "reduction of equipment

18
to accurate drawings, and results to well-written reports...."

Its goal was dual, and sounded simpler than it was: "to

19
produce and use computers." Systems engineering, Forrester

explained in his report prepared to meet Compton's requests,

involved "the knitting together of important and valuable

new systems from old and new components" in order to demon-

20
strate "the useful application of the research results."

In this assertion he was a trifle wide of Whirlwind's mark,

as a consequence of the spectacular lack of old components

and the hazy prospect of nothing but relatively formless and

untried mechanisms . It could be argued that vacuum tubes

and crystal diodes and circuits of all sorts were really just

"old components," but to those interested in the prospects

of electronic computers these were not the interesting or the

8.22
vital components, except to the engineer. The impressive
components were the computational "heart" of the machine and
its internal "memory," for with a sufficiency of these, ap-
propriately controlled and tied to information input and
output devices, a computer really became a computer worth
thinking about.

While Project Whirlwind sought a "systems approach" to
the building of computers, major interest elsewhere in the
nation continued to center upon questions of what performance
one might expect from a finished machine. From performance,
prowess could be estimated; the kind of performance in view
was calculational, and the kind of prowess esteemed was
logical and mathematical. Project Whirlwind, on the other
hand, was spending all its energies--and all those ONR dollars--
on prior questions of physical structure and electronic per-
formance, rather than on calculational performance. This
was a consequence of the fact that attention at MIT focused
on empirical considerations which the young engineers in the
Barta building considered inescapable. To them it was at
once a truism and a serious fact of engineering life that

"in many systems the greatest difficulties lie in achieving

21
the required reliability."

The Project leaders sharply appreciated and shared the

view that "producing a satisfactory working system often

requires greater technical contribution than producing the

22
basic components of that system." Engineering research

8.23
and development must be combined with system considerations;
this was a policy commitment ingrained in the very name of
the Servomechanisms Laboratory, and it was a policy commit-
ment the abrogation of which the Project leaders found un-
thinkable when trying to design and build the Whirlwind
computer .

As the authority of Special Devices Center personnel
faded and that of Mathematics Branch personnel grew, so the
visible respect that ONR felt for the IAS computer project
at Princeton became more significant. Aware of this trend,
Forrester and Everett sought to show how different was MIT's
systems-engineering approach from that Pursued by von Neumann,
Goldstine, and their associates. They realized, as has been
remarked at the beginning of this chapter, that the two
philosophies of research and development in question started
from different postulates and followed different routes in
reaching their common goal, the manufacture of a working
computer.

Any comparison based on adoption of either of these
philosophies as a standard could only judge one project at
the expense of the other and produce invidious comparisons
while hopelessly confusing and intermixing the differing means
and ends of the two projects. If the MIT project were selected
as the norm, then the IAS project must be considered inadequate
and unacceptable. If the IAS approach were to provide the cri-
teria, then the MIT procedures must be rejected as wasteful

8. 24-
and inappropriate .

Since von Neumann and Golds tine had made it abundantly
clear in 19M-6 how profound was their understanding of the
potential value of the automatic-sequence-controlled-
calculator mode of attacking hitherto prodigious and un-
assailable problems by mechanical (including electronic)
means, the two unheralded young MIT engineers were at a
disadvantage from the start. Nevertheless, they hammered
away at the differences between the IAS and the MIT research
and development procedures. "IAS," they pointed out in
their analysis to Compton in October, 19M-8, "is presently
engaged in constructing what is essentially a breadboard
model of a computer." MIT, on the other hand, "is building

what can more correctly be called a prototype and not an

23
experiment or a breadboard." This analysis of their dif-
ferences in 19M-8 was equally descriptive and to the point
in 1949 and 1950.

Fellow engineers, as well as basic-research scientists
and mathematicians pure and applied, could be expected to
perceive the distinction: IAS was committed to making one
of a kind, while MIT was fabricating the parent of a sub-
sequent line of computers. Obviously, the latter effort was
the more ambitious, since not a half-dozen computers had yet
been put into successful operation.

The experience to date was so limited and the field of
development was so wide and so full of unknown pitfalls of

8.25
all sorts that there was no way in the world of guaranteeing
in advance that the MIT venture would not come a cropper.
If what Everett and Forrester were saying was true, then the
enormity of the risks they ran was obvious to anyone who had
any sort of acquaintance with the problems of developing new
machines that must work when built. In this respect, the
apprehensions of ONR personnel that they might well be pouring
money down a bottomless though chromium-plated and gleaming
drain appeared well-grounded indeed; only after the fact
could they reliably take the measure of the MIT enterprise.
There was no way of knowing whether the Whirlwind approach
was catastrophically premature or a dramatic leap forward.
If the lessons of all past research and development ex-
perience were worth anything, they suggested that Project
Whirlwind would most likely turn out to be neither of these
alternatives. It would become instead an attenuated fizzle,
discreetly squelched, from which useful gleanings might be
garnered in such salvage operations as would prove practical
before the whole business was quietly swept under the rug of
the obligingly silent past.

Obviously, the Whirlwind group stoutly rejected such a
dismal prospect. After all, they knew what they were doing,
and in the intimate fullness of this knowledge, they explained
that "on the basis of considerable study, MIT has reached a
fairly firm conclusion as to the nature of the computer
needed." What was "fairly firm" supposed to mean? Was it

8.26
to become another funnel down which MIT would ask the Navy

to pour another million dollars when even a quarter of a mil-
lion would be risky and hard to come by?

Certainly there was no disagreement regarding the aptness
of Forrester's and Everett's admission that "much is still

to be learned that can only be learned from this machine

24
itself." At least, later investigators might prosper from

their mistakes made possible by de Florez's original enthu-
siasm for the aircraft analyzer and ONR's subsequent reluctant
expenditures. So might one reflect gloomily.

The MIT engineers went into amplifying detail, to in-
dicate how they hoped to avoid large mistakes (including the
production of a machine that would be obsolescent before it
was finished) . But these amplifications, designed to support
MIT's case, could as easily be read and as reasonably be
interpreted to raise new spectres for ONR, because the funda-
mental issue, undemonstrated by a thousand or a million words,
lay in the question of whether the talent to reduce to practice
in the manner they were proceeding was a talent the young
engineers—men not even of Ph.D. rank--really possessed or
not. Consequently, when they declared that their prototype
was being built "as near to the presently foreseen needed
characteristics as possible, with the following differences,"
the effects of these remarks could only be to reassure those,
such as Nat Sage, who remained confident of their abilities
and to redouble the misgivings of those such as Mina Rees,

8.27
who were uneasy yet knew they were responsible and who ex-
pected to be held to account.

The curious feature of these statements by Everett and
Forrester is that they became apt descriptions and accurate
forecasts of the procedures by which Project Whirlwind actually
carried forward its research and development investigations.
As descriptions and forecasts, they were idealistic in tone,
as is customary, and to the degree that they represented the
smooth, untroubled tenor of events in the world of the ideal,
they of course failed to recognize those rough edges of
reality that give the world of experience its relatively
scratchy character. The MIT engineers happened to possess
sufficient sense of proportion, however, to employ usable
ideals convertible to practical expression in the forging of
events that constitutes the research and development process.
Historically speaking, this judgment becomes easy to render
after the fact: had the young men failed, then ipso facto
they would have lacked the talent required; since they suc-
ceeded beyond even their own first dreams (although not
later ones that came with greater knowledge and experience) ,
then equally obviously they possessed the needed talent, and
the estimates and judgments they employed in gauging future
general needs were indeed appropriate.

"Great flexibility is being built in," they pointed out.
"Every facility for easy study, maintenance, and modification
is being provided. Wherever compromise on specifications has

8.28
been necessary it has been made only with provision for later
improvement and without relaxing the specifications for other
elements. Where necessary to meet specifications, special
component research has been undertaken. Elaborate and some-
times redundant trouble -location and prevention equipment
has been designed. The intention is that the prototype should
embody as many as possible of the desired features and

characteristics, and to insure this the prototype will probably

25
include many which are not needed."

This was how they expected to carry on the research and
development program they had long since begun. Furthermore,
they had adopted a strategy of research, design, build, and
test sharply different from that which von Neumann, Goldstine,
and their associates employed: the IAS approach to the
problem of building a machine unlike any yet built was ex-
perimental. To call it trial-and-error was to distort its
true character, for there was no blind casting about, no
"let's see what happens if--." It was a plan of attack that
shifted back and forth from the. realm of the ideal to that
of the practical, in order to see how close an approximation
could be obtained between the performance of physical equip-
ment and the execution of logical procedures.

The IAS builders would be willing to go back to the
drawing boards more times than would the MIT builders in
order to achieve a given degree of technical improvement.
The IAS approach, said Forrester and Everett, was "to attain

8.29
the desired goal by an iterative procedure, the first step
of which is a single attack aimed in the estimated direction."
This attack would produce a test-bench, or "breadboard,"
device, the subsequent performance of which would tell them
whether it needed refinement or whether it would suffice
until such time as connecting it up with other elements in a
system would reveal def iciencies--a frankly and honestly
linear and experimental procedure .

Project Whirlwind's approach was more ambitious: "to

estimate the goal more exactly and then to flood or saturate

27
the area surrounding that estimate in a complex attack."

Although subsequent fiscal, administrative, and programming
events were to show that these words fell on deaf ears and
that even ears at MIT seemed to grow slightly hard of hearing
at times, this description of the research and development
approach the Project was following was quite honest and
accurate. Indeed, it was a strategy the engineers had de-
liberately adopted and adhered to, not fallen into, as a
consequence of their World War II engineering experience in
Professor Brown's Servomechanisms Laboratory.

Unfortunately, its virtues were not apparent, even though
it was a technique of procedure superbly fitted to cope with
certain problems inherent in the analyses of systems of
machinery. It was expensive. It was elaborate. It was not
widely used, partly because it was so expensive, partly be-
cause it placed such unremitting emphasis upon premium- quality

8.30
performance, and partly because it required a rare measure
of engineering sophistication, experience, and insight. In
addition, those most interested in the new computers and most
influential were not versed in such engineering modes of
procedure; they were interested in scientific problems, in
the tantalizing, potential applicability of the computer,
in mathematical problems, or in information-retrieval problems,
and because these interests were non-engineering in their
direction and did not join issue on the policy level with
the problems of design, fabrication, and performance, they
failed to appreciate the power, the virtue, and the relative,
long-run cheapness of such a formidable and, in a sense,
daring research and development procedure.

Thus, the Project and its way of doing things were vul-
nerable not only because of the relatively small funds made
available by ONR for fiscal year 1950--three-quarters of a
million dollars--but also because of the peculiar, if not
unique, and costly nature of the research and development
procedures that the MIT engineers insisted on adhering to,
so different from traditional and prevailing modes accepted
by Navy administrators during the late Forties.

NOTES TO CHAPTER 8

1. Jay W. Forrester, Memorandum L-5, "Project Whirlwind,
Principles Governing Servomechanisms Laboratory Research
and Development," Oct. 11, 1948, cc to: Dr. K. T.
Compton, Dr. J. R. Killian, Mr. Henry Loomis, Mr. N.
McL. Sage, Dr. G. S. Brown, Mr. Ralph Booth. The memo,
one of the "L" -series of Whirlwind reports, covered
three typewritten pages, single-spaced.

2. Jay W. Forrester and Robert R. Everett, Memorandum L-6,
"Comparison between the Computer Programs at the Institute
for Advanced Study and the MIT Servomechanisms Laboratory,"
Oct. 11, 1948, cc to: President's Office (2), Mr. N.

McL. Sage, Prof. G. S. Brown, Mr. Ralph Booth. This
"L"-series report covered six typewritten pages, single-
spaced.

3. Memorandum L-5, p. 1.

4. Memorandum L-6, pp. 2, 5.

5. Memorandum, Hugh R. Boyd and Robert A. Nelson to Jay W.
Forrester, subj . : "Detailed Estimates of Costs Project
Whirlwind, Applicable to Period from November 1948
through June 1950," November 4, 1948; Memorandum L-8 ,
J. W. Forrester to N. McL. Sage, subj.: "Cost Trends,
Contract NSori-60," November 19, 1948; Memorandum, J.

W. Forrester to N. McL. Sage, subj.: "Budget Adjustments,"
December, 1948.

6. J. W. Forrester, Computation Book No. 45 , entries for
November 2, 1948, p. 141; December 6, 1948, p. 146.

7. J. W. Forrester, Computation Book No. 49, entries for
December 9, 1948.

8. Amendment No. 7, T. 0. No. 1, Contract N5ori-60,
March 31, 1949.

9. Ltr., N. McL. Sage to ONR, December 10, 1948; ltr.,

Mina Rees to N. McL. Sage, February 23, 1949; Memorandum,
Harris Fahnestock to J. W. Forrester, subj.: "Comments
on Letter Dated February 23, 1949 from Dr. Rees to Mr.
Sage," March 15, 1949. J. W. Forrester, Computation Book
No^ 49, entries for 1/12/49 and 1/13/49, pp. 15-17. See
also Memorandum L-10 (Draft), J. W. Forrester to N. McL.
Sage, subj.: "Analysis of Whirlwind Program," January
13, 1949.

10. Ltr., J. W. Forrester to C. V. L. Smith, May 20, 1949;
ltr., N. McL. Sage to Head, Computer Branch, Mathematical
Science Division, ONR, June 14, 1949; ltr., C. V. L.
Smith to N. McL. Sage, July 18, 1949; Amendment No. 9,

T. 0. No. 1, Contract N5ori-60, July 1, 1949; ltr. M.
S. Stevens to Head, Computer Branch, sub j . : "Supplementa-
ry information on request for funds for Contract NSori-60
(NR 048 097)," July 27, 1949; File Memorandum, J. W.
Forrester, sub j . : "Air Traffic Control Project," March
10, 1949; J. W. Forrester, Computation Book No . 45 ,
entry for July 28, 1948, p. 121; Computation Book No. 49,
entry for March 10, 1949, p. 36.

11. J. W. Forrester, Memo L-10 (Draft), subj . : "Analysis of
Whirlwind Program," Jan. 13, 1949, p. 1.

12. Summary Report No. 15, Dec. 1948, p. 2.

13. Ibid .

14. Summary Report No. 16 (Jan.) , No^ 17 (Feb.) , No_;_ 18
(March 1949) .

15. Summary Report No. 18, Mar. 1949, pp. 14-15.

16. The engineers interviewed here by the authors requested
their identities remain anonymous.

17. Memorandum L-S , subj.: "Project Whirlwind, Principles
governing Servomech. Lab. Research and Development,"
October 11, 1948, p. 2.

18. Ibid .

19. Memorandum L-6, subj.: "Comparison between the Computer
Programs at the Institute for Advanced Study and the MIT
Servomechanisms Laboratory," October 11, 1948, p. 2.

20.

Memorandum L-5,

p. 1.

21.

Memorandum L-5,

p. 2.

22.

Ibid., p. 1.

23.

Memorandum L-6,

pp. 1, 2.

24.

Memorandum L-6,

p. 2.

25.

Ibid.

26.

Ibid.

27.

Ibid.

CHAPTER NINE
THE COLLISION COURSE OF ONR AND WW

When the Office of Naval Research in the spring
of 1949 made its conservative allocation of fiscal
year 1950 Whirlwind funds, a skirmish or even a battle
may have been lost, but Forrester had not yielded
the field. In December, 1949, responding to a re-
quest from the deputy Director of ONR, Captain J.B.
Pearson, Forrester projected into fiscal years 1951
and 1952 a program for digital computer work at MIT
which would have cost $1,150,000 and $943,000, re-
spectively. The program he envisaged was quite ex-
pensive, including a "normal continuation" of the
existing program and an expanded program for research
in the area of application. Again Forrester warned
against "the over-optimism and unfounded promises
which have been so apparent in much of the digital
computer planning and publicity. " The programs would
be long, and sponsors could not expect "immediately
hardware for the complete solutions of their own
problems . "

Forrester's response to Captain Pearson brought
forth some rather strong opposition from various

9.1

9.2
members of ONR. R.J. Bergemann, Physical Scientist
for the Boston Branch Office, Severely attacked
Forrester for not containing his program within the
limits established by ONR. Instead of planning to
complete the computer at minimum cost, he charged,
Forrester's thinking was directed "towards the great
possibilities that lie in computer application."
Herein lay Forrester's sin, for he had clearly been
instructed, according to Bergemann, to eliminate
"long range planning." Bergemann directly attacked
the Project for producing "less for the money than
might be obtained elsewhere , " and he unfavorably com-
pared it to the "Hurricane" computer under development
at Raytheon.

The Raytheon project, Bergemann argued, was tech-
nically superior and cost less, primarily because the
men engaged in it possessed greater experience and
competence. Whirlwind personnel on the other hand, he
noted, had had no "previous digital computer experience,"
and few of the Project's engineers had had "any engineer-
ing experience other than under OSRD-NDRC contracts
where cost was no object."

Bergemann' s recommendation was that "ONR reemphasize
the necessity for lower expenditures in Project Whirlwind
by concentration of effort on completion of the computer

9.3
in its simplest useful form." Forrester, he explained,
must be made aware of the difficulty of justifying the
spending of "one twentieth of the ONR budget on his pro-
ject, when Raytheon has done so much more on a smaller
expenditure. " 2 Unfortunately for Bergemann, the project to
which he so unfavorably compared Whirlwind did not measure
up to expectations. Within the year the recommendation
was made that the Raytheon contract be terminated, upon
the grounds the company could not meet its "contractual
obligations ... with their existing organization, on their
presently estimated schedule and at the estimated cost." 3

C.V.L. Smith, Head of the Computer Branch, was another
who seriously attacked Forrester and his Project, referring
to the latter' s estimates in his letter to Pearson as "fan-
tastic." He found "appalling" Forrester's refusal to re-
cognize "that funds simply are not available to support
such an extensive program." Smith also found the program
projected for Whirlwind "excessive" and the staff not
sufficiently qualified "to justify this expenditure." He
did not, however, repeat Bergemann' s unfortunate mistake
of comparing it unfavorably to the project under develop-
ment at Raytheon. He summed up his argument by recommend-
ing that Whirlwind be made operational during 1951 and that
Forrester be convinced of the necessity to reduce drastically
expenditures and to stop thinking "in terms of a million
or so per year."

9.4

In response to a proposal, apparently advanced by
Pearson, that a conference be called to discuss the
financing of Project Whirlwind, Smith's attitude was
negative. He opined that Whirlwind had been "oversold,"
that "a very considerable skepticism" had arisen. It
would be "a great mistake" to call a meeting before it was
possible to demonstrate a fully operable machine." He
suggested the machine be tested by running several diverse
problems on it to permit "a really convincing demonstration"
of its potential. "Anything short of this would not only
be futile, but probably harmful in its total effect."

It is interesting to note that during the course of
a visit to Project Whirlwind on January 12 and 13, neither
Bergmann nor Smith was, understandably, as caustic in his
comments to Forrester and his colleagues as each permitted
himself to be in memoranda intended for internal Navy eyes.
At least, the trip reports prepared by the two men give no
evidence of such blunt and candid exchange. Smith did,
however, upon this occasion review with Forrester the latter 's
proposed budget for fiscal year 1951, explaining that it was
impossible for ONR to raise the 1.15 million dollars pro-
posed and that at best the office was planning to allocate
$250,000 to $300, 000. 5

Once again Forrester's proposals on program and budget

9.5
for Project Whirlwind raised the matter to the highest
levels within both MIT and ONR; once again the decision
was taken to discuss the matter in a general conference to
be hosted this time by the Institute; and once again Forrester,
in preparation for the exchange of views , sought to win the
Institute's administration to his side. In a letter to the
Provost, Dr. J. A. Stratton, Forrester explained in great
detail the! capabilities of the computer which, he wrote,
would be assembled by the fall of 1950 and ready "to start
research into 'real-time* applications." He predicted the
computer would be capable of "preliminary work" in at least
eighty per cent of the applications listed in his letter
to Pearson, including fire-control studies, logistics studies,
centralized digital computer service, weapon evaluation,
engineering and scientific applications, antisubmarine-
control studies, air-intercept combat-information center
research, simulation, and air-traffic control research. 6

This time, however, Forrester was less successful in
persuading his superiors to give him full support. Viewed
from one direction, the Navy, or more particularly, ONR,
finally was able to execute an end-run around Forrester and
Sage and reach MIT's top administration without effective
interference. Viewed from another direction, Forrester had
been unable longer to convince his superiors that he re-

9.6
cognized the funding realities of life that prevail in
the conduct of research and development during that un-
eventful spring before the Korean War suddenly broke out.

From still another direction, one might speculate
that Whirlwind had become a computer without a practical,
specified mission as a consequence of Forrester and
Everett's commitment to an avowedly general-purpose in-
strument, and as a consequence of their single-minded
concern to bend all their efforts to bringing such a com-
puter into being. The research and development process
itself had for years demonstrated the practical worth of
Benjamin Franklin's shrewd rhetorical query (uttered in
reply to a critic of the first balloon flights) — "Of
what use is a new-born baby?" — but this general wisdom
did not automatically justify, in the particular instance
of the Whirlwind project, the torrential outpouring of
ONR dollars that Forrester sought. Everything in the
traditional philosophy of government funding of research
and development indicated that the Navy could not afford
to gestate so costly a baby of so uncertain pedigree when
more promising purebreds, such as the IAS machine, were
costing so much less. Nor did it simplify the problem to
have some of the Navy program managers feel that Forrester's
reiteration of his demands was bordering on the arrogant.

9.7
Forrester recalled in later years that he had shared the
apprehensions of Special Devices Division personnel re-
garding confidential projections which called for a Russian
atomic strike capability by 1953. 7 These concerns were less
central in the minds of the mathematics and science orient-
ed programmers in ONR who were responsible for maintaining
liaison and surveillance relations with Project Whirlwind
by early 1950.

Finally, it could be argued from still another direc-
tion that the series of investigations and inspection-
visits instituted by MIT and by the Navy over the past three
years were by this time exerting an appreciable cumulative
effect. In any event, these investigations and their find-
ings came to constitute a factor that did not improve —
if it did not actually harm — Project Whirlwind's chances
of gaining and maintaining the degree of financial support
Forrester so unremittingly and unrepentantly sought.

The determined efforts made by ONR to reduce the costs
of Project Whirlwind and to restrain Forrester indicated that
the early apprehensions of those directly responsible for
the administration of the Project had not been allayed. In-
stead, their concern eventually had reached even the highest
levels of ONR and MIT. Both Warren Weaver and Francis J.
Murray in 1947 had been relatively mild in their respective
evaluations, neither one finding any major flaws in the

9.8
Whirlwind program. Yet both had confirmed, if only mildly,
the fears of the Mathematics Branch of ONR that the Project
was weak in mathematical competence and direction. The
only evaluation which had been outspoken in its praise was
that prepared by Ralph Booth, assisted by J. Curry Street.
To some extent this investigation could be discounted, on
the grounds that it had been conducted by a sympathetic and
prejudiced investigator.

Other evaluations, however, were more sharply critical
if not outright condemnatory of the Project, its cost and
the absence of a definite purpose. One of these evaluations
was prepared sometime during 1948 for Mina Rees and without
doubt by someone in whom she had implicit confidence. His
criticisms were so strong that even sixteen years later she
declined to reveal his identity. This anonymous critic
sharply challenged the Project, finding it completely "un-
sound on the mathematical side" and "grossly over-complicated
technically." It was a program without purpose, one which
had become "one of the most ambitious in the country . . .
notable for the lavishness of its staff and building."
Apparently, there was little about the program and its dir-
ectors which the critic could praise, although he did, per-
haps grudgingly, approve Forrester's "ideas about great re-
liability and the necessity of convenient and complete pro-

o

visions for checking and for locating trouble..."

9.9
Acutely aware of the controversy revolving around
Project Whirlwind, of her own lack of understanding of
the engineer's approach, and of the necessity for a valid,
competent, and objective evaluation if her recommendations
concerning the program were to possess substance and merit,
Mina Rees undertook in late 1948 and early 1949 to organize
an inquiry which would at one and the same time familiarize
her with the program and provide the critical analysis she
needed. The organization of a team for this purpose was not
easy, for although she could appoint members from the ONR
staff, it was exceedingly difficult for her to find an im-
partial expert acceptable both to her and to the administra-
tors of MIT and the Project. 9

Eventually, Dr. Harry Nyquist of the Bell Telephone
Laboratories was settled upon. The committee, comprised of
Dr. Nyquist as the impartial expert, Mina Rees, C.V.L. Smith,
and Dr. Karl Spangenburg, head of ONR's Electronics Branch,
visited the Project in the spring of 1949. The group review-
ed and analyzed the program, finding apparently no major
weaknesses. Some technical questions were raised regarding
"the means of communicating with the machine," the "means of
auxiliary storage , " the computer ' s word-length and the
storage tube development program, but all in all the group,
according to C.V.L. Smith, "was favorably impressed by the
thoroughness of the engineering effort displayed by the

9.10
Whirlwind staff, and by the energy, enthusiasm and direct-
ness of approach with which the numerous difficult problems
encountered have been attacked." 10

In response, Forrester expressed his appreciation. At
the same time, he noted that in an earlier communication he
had anticipated the committee's recommendations by suggest-
ing new task orders to cover the proposed work. 11 This was
not exactly what the committee had really recommended, for
new task orders meant additional funds.

The evaluation which hurt most came at the end of 1949,
and it brought a sharp and irritated rejoinder from Forrester.
This was the investigation which the Chief of Naval Research
had been anticipating, conducted by the "Ad Hoc Panel on
Electronic Digital Computers" of the "Committee on Basic
Physical Sciences" of the Research and Development Board.
The Ad Hoc Panel had been created on July 29, 1949. It was
composed of Dr. Lyman R. Fink, chairman, Dr. Gervais W. Trichel,
and Dr. Harry Nyquist, and it proposed "to look critically at
the several projects comprising the program on digital com-
puting devices in the Department of Defense, with emphasis
on the objectives, management, engineering planning, current
status, and probability of successful completion of each
project. "

After visiting various contemporary digital computer

9.11
projects, holding hearings, attending the Second Symposium

on Large Scale Digital Calculating Machinery at Harvard
University and studying contemporary progress and engineer-
ing reports, the committee prepared and issued a tentative
report of its findings and recommendations on December 1, 1949.
Jay Forrester, it is interesting to note, was not included
in the distribution list; son Neumann was.

The panel concerned itself with the total computer pro-
gram supported by agencies within the Department of Defense.
Forrester, in his rebuttal, observed, however, that the panel
had overlooked the United States Naval Computing Laboratory
operated in St. Paul, Minnesota by Engineering Research
Associates on a budget and staff level three times that of
Project Whirlwind. This was, perhaps, his retort to the
panel's conclusion that the "scale of effort" on Project
Whirlwind was "out of all proportion to the effort being ex-
pended on other projects having better specified objectives."
If the panel's figures were even approximately accurate,
Project Whirlwind's estimated completion costs, made at a-
bout the same time as the panel was conducting its inquiry,
were about twenty- seven per cent of the total amount the
panel estimated would be the cost of the overall Department
of Defense program. This overall program, comprising some
thirteen machines under development by eight suppliers,

9.12
would cost, the panel estimated, some ten millions of
dollars.

Its tentative report was circulated for "information
and comment." In it the panel discussed broadly the need
for high speed digital computers , the requirements which
should be met for a rational, over-all program, and the
status of the contemporary program. In critically evalu-
ating the contemporary overall program and the individual
projects-in-being composing it, the panel found it to lack
coordination, organization, and centralization, and noted
that it failed, therefore, to realize "optimum" return from
the effort and money expended. The panel did not ascribe
these basic flaws to any particular agency or cause; in-
deed, it conceded that the projects appraised antedated the
establishment of the Department of Defense. The services
should in fact be "commended rather than criticised for the
degree of voluntary cooperation" that had taken place.
These words were probably the kindest the panel wrote into
the report.

Critically, the panel observed that "no specific pro-
cedure" had been established "for the review, coordination
and control of high speed digital computer development."
The over-all program was marked by the absence of a central
agency which could collect and distribute information or one

9.13
which could evaluate performance in order to provide new

users with "reliable sources for advice and assistance

in technical procurement in an unfamiliar field."

The panel further concluded that the technical guid-
ance and supervision provided the individual projects by
their respective sponsoring agencies were insufficient. In
several instances technical reporting was poor and not kept
current; technical directors were not always aware of the
contemporary state-of-the-art; the exchange of information
was often poor; contractors were not always given proper
direction; in some instances contractors had made "im-
portant changes in the operating characteristics of systems"
without approval; estimated dates of completion were not
realistic; and some devices were being "built as part of
contracts for other devices or incident to service contracts."

To reiterate: the over-all program lacked coordination,
organization, supervision, and centralized control. To cope
with and eliminate these fundamental weaknesses, the Ad Hoc
Panel recommended the creation by the Research and Develop-
ment Board of a panel subsumed not under a mission-oriented
engineering committee but under a committee that would be
expected to regard the computer for the scientific engine
the panel knew it to be. Such a panel should properly be
placed under the Committee on Basic Physical Sciences, to
coordinate the Department of Defense digital computer

9.14
programs. Any project not approved by the proposed panel
would be denied budgetary support. The existing projects,
however, the Ad Hoc Panel recommended should be left "sub-
stantially intact, since this would serve the "best inter-
ests" of the Department.

In its treatment of specific projects, the panel was
no more charitable than it had been in its treatment of
the over-all Department of Defense program; even the pro-
gram at the Institute for Advanced Study received its fair
share of critical comment. Project Whirlwind, although
commended for its "excellent job training" of graduate
students, for the excellence of its engineering and scien-
tific staff, and for the quality and quantity of its en-
gineering reports, was held to be lacking a "suitable end
use." Consequently the recommendation was made that if the
Navy could not find one, "further expenditure for the com-
pletion of the machine should be stopped." However, the
panel did suggest that consideration be given to using MIT's
excellent staff "on system studies and on the development
of specific computing components," especially the storage
tube. 12

The reaction of Forrester and his colleagues to the
panel's findings and recommendations was one of distress,
concern, and anger. Acknowledging many of the panel's gen-
eral recommendations to be excellent, they opined that the

9.15
portions of the report which dealt with specific projects
were incomplete, based upon inaccurate information, and
superficial. The panel had stressed the flaws and weak-
nesses disclosed by the investigation to such an extent,
Forrester charged, that it had raised the "real danger" of
"shaking confidence in the field" and destroying thereby

the efficacy of the general recommendations offered "for

13
strengthening the digital computer program."

Without doubt, a good portion of Forrester's irritation
and that of his associates found its genesis in the panel's
comments on Proiect Whirlwind itself. This is understand-
able, for the Whirlwind staff had prepared long and detail-
ed explanatory memoranda, describing the purpose and nature
of the program, its historical background, and the contri-
bution the Project was making and would continue to make to

14
computer technology — all apparently for nought.

In addition, many of the criticisms and recommendations
expressed within the report either reflected or were modi-
fied versions of the very points Forrester had been advancing
over the course of the preceding years. Without doubt, the
engineers and scientists of Project Whirlwind felt both let
down over what to them must have been the panel's failure
to recognize their contribution to the state-of-the-art, and
angered by the panel's recommendation that the program to
which they had dedicated themselves be eliminated unless a

9.16
specific and positive use for Whirlwind was found.

The cries of anguish and anger were not Project
Whirlwind's alone. The charges made by the panel were
sufficiently penetrating to compel the Acting Head of the
Computer Branch of ONR, A. E. Smith, to prepare a rebuttal
which in essence defended not only ONR, but Project Whirl-
wind as well. Replying to the specific charge that ONR
had no purpose in view for Whirlwind, Smith noted that when
completed, the computer would be "useful ... to point the
way to the solutions of the numerous control and real time
simulation problems of importance to the Department of De-
fense." Listing the various areas of application which had
been considered or which were under study at the time of
writing, Smith argued that each would require "voluminous
arithmetic experimentation . . . before goals can be set with
any precision or efficiency." This, coupled with the "in-
terest of the many different activities in Whirlwind," pro-
vided justification enough to proceed with the program. As
far as Smith was concerned, Whirlwind's completion was man-
datory, in order to realize "the original goal of the project"
and also the proposals made by the panel itself concerning
the use of the MIT group for system studies and the develop-
ment of components .

It was in such an atmosphere of investigation, criticism,
complaint, and counter-complaint, or in the climate influenced

9.17
in part by such developments , that Forrester late in Feb-
ruary and early in March, 1950, was unable to persuade his
superiors at MIT to give him the full support he desired.
One mighti.argue that the common cause allying ONR and MIT
against a common, hostile critic, the Research and Develop-
ment Board and its Ad Hoc Panel, caused the two organizations
under attack to submerge their smaller differences, such as
how much funding support to give Project Whirlwind, and ex-
pediently to close ranks to deal with the issue at hand.
Whatever the combination of causes , Forrester found himself
corralled as never before. Three days after his letter to
Stratton, ONR representatives came to Cambridge to discuss
the Project and its place in the overall ONR computer pro-
gram. Two conferences took place on March 6, 1950, one in
the morning and one in the afternoon. At the morning con-
ference were Provost Stratton, Dean Harrison, and Dean Sher-
wood of MIT, and Dr. A. T. Waterman, Dr. Mina Rees, and Dr.
C. F. Muckenhaupt of ONR. This was a policy meeting which
Forrester did not attend, and this fact suggests not only
did the MIT administration recognize that some of the Navy

criticism of Project Whirlwind was taking strong colors of personal criticism
of Forrester's way of conducting his affairs, such a thorn had he become, but
also the MIT authorities recognized the wisdom of maintaining ONR support
in common cause against such Defense Department criticisms as the

9.18
Ad Hoc Panel had levelled. It was to the advantage of
both the parties to resolve existing differences as un-
emotionally as possible.

In any event, Forrester did not attend the meeting.
It appeared that the MIT leadership had decided to formu-
late, with ONR representatives, broad guidance principles
to which Forrester and his associates would have to con-
form. But the MIT leaders could afford to pursue their
course tactfully and magnanimously, for they had as ace
up their sleeve, and of this circumstance Forrester was
reassuringly aware, especially since the Project had
assisted in placing it there.

The morning conferees discussed MIT's thoughts con-
cerning the "advisability" of combining the Institute's
computer programs under a single head while permitting the
individual programs to continue within the departmental
structure. The representatives of ONR welcomed the pro-
posed reorganization, provided a "suitable head of the
program ... be chosen. " Forrester was not among those
considered for the appointment. The conference ended
with MIT's agreeing to see that Project Whirlwind lived

within a $250,000 budget for the following year, the max-

17
imum amount ONR could allocate to the program. Thus

Forrester's hopes appeared to be frustrated without any

possibility of an appeal to a sympathetic MIT administration.

9.19
Mention has been made that MIT had an ace up its sleeve.
Events suggest the Navy knew it was there and was equally
good-humored about it. The afternoon meeting was attend-
ed by all the morning conferees except Dean Harrison;
present also were Jay Forrester, C. V. L. Smith, and a new
figure, Dr. George Valley of the MIT Physics Department.
Valley was chairman of an Air Force committee that had
been created to investigate the state of contemporary air
defense with the purpose of recommending improvements and
changes. It was Valley's presence that altered the whole
financial picture for Project Whirlwind, for he proposed at
the aftern©on conference that Whirlwind be applied to ex-
periments in air defense. To this purpose, he believed,
the Air Force would be willing to allocate some $500,000.
All agreed that this would be an excellent solution to the
situation, for it would assist the Air Force in a problem

of great national importance, yet leave the computer avail-

18
able for scientific use and for use on Navy problems also.

The following day, March 7th, another meeting was held

to discuss in greater detail the financing of Project

Whirlwind and the program suggested in Forrester's letter

to Captain Pearson. The financial basis for the discussion

was $300,000 from the Computer Branch of ONR plus another

$30,000 from the Armament Branch for a fire-control study.

Monies over and above this total of $330,000 would have to

9.20

come from sources other than ONR, in this instance from
the Air Force for its air-defense study.

The two-day conference did demonstrate that neither
MIT nor ONR was inclined to permit Project Whirlwind to
become an operation resembling in size or cost such efforts
as the wartime radar program or the Manhatten Project, as
Forrester had occasionally suggested. If it had not been
for the Air Force and its search for an adequate air de-
fense system, Project Whirlwind might well have been limit-
ed to scientific calculations and such modest Navy projects
as might have arisen. The program underwritten by ONR a-
lone would never have met Forrester's desires or expectations,
In the imposition of limitations upon the program, ONR had
finally won the cooperation of MIT, aided by the fortuitous

Of)

cooperation of the Air Force.

Accepting $780,000 as the maximum allocation for fiscal
year 1951, Forrester planned the Project's program accord-
ingly. He submitted a memorandum to this effect to Nat Sage,
who in turn forwarded it to ONR in May of 1950 as an en-
closure to the Institute's official request for funds for

fiscal year 1951. In addition to the $780,000 $280,000

from ONR and $500,000 from the Air Force Forrester an-
ticipated an additional $120,000 from the Air Force for the
Air Traffic Control study and approximately $32,000 for an
additional Navy study in the application of digital com-
puters to fire control: a total of $932 ,000 .^

9.21
Late the following month C. V. L. Smith, Head of the
Computer Branch, replied that scientific approval of the
proposed budget had been granted and that MIT "would
shortly hear from ONR's Contract Division. On June 29,
1950 the Amendment officially confirming the allocation and
extending the time of the contract to June 30, 1951 was
issued. In his letter, Smith remarked that it was planned
by ONR that the $280,000 would carry the Project for about
four and one half months; meanwhile, the Air Force, he
anticipated, would transfer to ONR $500,000 of Air Force
Funds for fiscal year 1951, to carry the Project to June
30, 1951. 22

In a comment to Nat Sage, Forrester observed that the
ONR allocation was for a four and one-half month period,
and it was his intention to implement his program on the
assumption an additional $500,000 would be forthcoming from
the Air Force to finance the program for the balance of the
fiscal year. Sage's brief reply that Smith's letter of
June 26 provided the answer to Forrester's implied question
suggests that Sage had either acceptted the impossibility of
obtaining more funds from ONR or had become convinced that
the Air Force would accept the recommendations of George

Valley and his committee and underwrite the proposed ex-

23
penmental program in axr defense.

9.22
The Air Force was slow in making its funds avail-
able, however, perhaps in part because of the indecision
and confusion which resulted from the intensification of
concern over the state and adequacy of the nation's air
defenses. This concern was a product of the Cold War,
which had become more ominous with Russia's detonation
of an atomic bomb in August, 1949, and also a product
of the outbreak of active fighting which occurred when
Communist North Korea invaded the United States-sup-
ported Republic of Korea to the south in June of 1950.
Finally, in mid-November, 1950, the Air Force transferred

to ONR $480,000, to which the Navy added $20,000 to pro-

24
vide the anticipated $500,000 for the air-defense study.

The entry of the Air Force into Project Whirlwind,
together with reorganizations carried out by MIT to cen-
tralize and coordinate computer-development projects in
which various of its faculty were engaged, resulted in
a broadening of the Institute's involvement with digital
computers and furnished a not uncommon instance of the
growing importance of the computer on the American tech-
nical scene at that time. In an action that left the
Whirlwind project freer to pursue its air defense inves-
tigations and that was carried out in part at Forrester's
suggestion, MIT ultimately established a Center for
Machine Computation under the direction of Professor
Philip M. Morse of the Physics Department. His became

9.23
the responsibility to "combine and coordinate the use of
existing computing machines at the Institute," both In-
stitute and government owned, including Whirlwind, but

25
Morse's Center was not concerned with air defense problems.

Formally, the Whirlwind staff that originated in the old

Servomechanisms Laboratory of the World War became the

Digital Computer Laboratory already housed in the Barta

Building, under the direction of Jay W. Forrester, with

Professor Gordon Brown as Faculty Adviser, Robert Everett

as Associate Director, Harris Fahnstock as Executive

Officer, and J. C. Proctor as Personnel and Security

Officer. 26

The Whirlwind group eventually was to join other MIT
groups and become incorporated, as "Division Six," into
the Lincoln Laboratory, which was established by MIT to
research, design, and develop a centralized air defense
network utilizing a high-speed digital computer at its
center. The Whirlwind computer was to play a dramatically
important role in demonstrating the feasibility of such a
system. It was also to be used by the Navy, which con-
tinued to allocate rather substantial sums of money to
MIT for research and development in the mathematics of
computer design and use.

Beginning with fiscal year 1951, although still carried
under the original contract, allocations were designated

9.24
by ONR for use not by Project Whirlwind, but rather by the
Center for Machine Computation for research in applied
mathematics and for research which would lead to improve-
ments in computers , and which would advance the state-of-
the-art. To this end, the Center was allocated $600,000
in June of 1951, and an additional and final allocation

of $50,000 was granted for studies in the application of

27
the digital computer to air defense. In March of 1952

$250,000 was made available; a year later $285,000. Al-
though subsequent allocations were made, they were much

less; these apparently were the final allocations of sub-

28
stance to be made by ONR under Contract N5ori-60.

The success by March of 1950 of ONR's policy to re-
duce the dollar cost of Project Whirlwind to the Navy
does not appear to have -been contingent upon Air Force
willingness to follow Professor Valley's judgment and
"pick up the tab," nor was it made possible by the fact
that Whirlwind I finally appeared to have found the prac-
tical reason-for-being (after long ago giving up the air-
craft simulator) that the Ad Hoc Panel had criticized it
for lacking. The principal reason may have been the Nasry's
willingness to write off to experience the unacceptably
high cost of maintaining Project Whirlwind's lavish stand-
ards of operation that temporarily had prevailed, for ONR
never did endorse Forrester's mode of operation, and

9.25
neither did the Air Force or the Department of Defense.
But the situation was probably not that simple.

It could be argued that, in a sense, even MIT found
it necessary to repudiate Forrester's way of doing things,
as Nat Sage's letter of July 11,1950 implied. It could
be argued further that Forrester's style could not be MIT's
style, under the circumstances of limited funding that
prevailed on the federal government level before the Korean
War broke out in June. Even then, there was a delay while
government echelons convulsively executed about-faces. So,
one might conclude, it was not the essential merit of the
Project's research and development record and performance
that moved the Air Force to follow Valley's lead, but rather
it was the emergency, "crash-program" nature of the need
imposed by the nation's vulnerability to aerial attack from
over the Pole that furnished Forrester his reprieve.

There remains still to be taken into account the be-
havior of the Navy and its ONR program managers. As Forrester
recalled, in spite of their repeated protests the Navy mana-
gers did continue to support the program financially during

29
the crucial development years of 1947, 1948, and 1949.

When the manager's misgivings grew sufficiently intense, an
investigating committee made a dubious or negative report,
this disturbing news was sufficient to generate the appoint-
ment of another investigating committee. In short, if the

9.26
Navy did not dare — or did not care — to support wholeheartedly
the Whirlwind project, with its insatiable funding demands,
neither could it shut it off, apparently.

ONR's growing reputation with the scientific community
was not likely to be shattered or even seriously sullied by
a simple decision to stop funding an unorthodox engineering
venture preoccupied with a scientific machine and directed
by young men who were still scarcely more than graduate
students and considerably less than professional faculty.
What about pressures within the Navy, then? They were not
all one-sided. Administrators in Mina Rees's position, for
example, could be expected to respond to the multiple pres-
sures that existed within their own organization, yet these
pressures did not appear to generate an obvious and un-
varying resultant whose measure can be quickly taken in
retrospect. The military establishment was looking closely
into computers when these were still largely disembodied
and tantalizing promises . It became a continuing military
commitment, nevertheless, even when an agreed-upon military
mission was not discernible, and the cross-currents of
technical promise and practical doubt could as well have
made it awkward as have made it easy for ONR to drop Whirl-
wind. Yet, on balance it appears to have been awkward
rather than easy.

As the months passed into years, the R&D momentum

9.27
which the Project built up became a factor contributing to
its survival. While the technical, visionary, command-and-
control projections that Forrester and Everett had committed
to paper in their early L-l and L-2 reports on antisubmarine
warfare techniques failed to impress the Navy with the ob-
viousness of their anticipation of things to come, neither
did the mounting dollar costs of the Project cause Admiral
Solberg, Alan Waterman, Manny Piore, Mina Rees, C.V.L. Smith,
and their associates to agree to "pull the plug".

The reticent character of the recollections of some of
the Navy programmers in after years may prompt the disin-
terested observer to infer that they had the well-known
bear by the tail, were unable to let go, and would rather
not remember it, but the fact remains that in those pioneer-
ing days in the history of computers they were extending con-
tinuing financial support with the palm of one hand while
extending continuing technical skepticism with the back of
the other. Perhaps the soundest lesson to be offered here is the
typical character of the R&D situation in which the MIT engin-
eers and the ONR programmers found themselves. Although the par-
ticular events were of course unique, the pattern of relations
between buyer and seller was customary. Those familiar with the
conduct of R & D will recognize that Project Whirlwind was not
staggering from precipice brink, as in some melodrama, but was
following, with respect to the events set forth here, a conven-
tional course in its pursuit of R & D.

9.28
The ambivalence of the Department of Defense with re-
gard to Project Whirlwind's situation has already been
indicated in the account of the preliminary report of the
Ad Hoc Panel to the Research and Development Board. Re-
affirmation of this dissatisfaction, slighly muted, is to
be seen in the final report of June 15, 1950, even though
by this time the status of Whirlwind had been considerably
altered by the Air Force's decision to accept the recommenda-
tions of the committee headed by Professor George Valley and
to employ Whirlwind to test the feasibility of establishing

a centrally controlled air defense system with a high speed

30
digital computer as its nucleus. Jay Forrester did not

neglect to bring this decision to the attention of the Re-
search and Development Board, which had authorized further
hearings and discussions to be held in order to air the var-
ious criticisms of the tentative report of December 1, 1949,
so that proper corrections and modifications could be made. 31

The Ad Hoc Panel's parent Committee on Basic Physical
Sciences had met on February 9, 1950, to consider the panel's
first report, submitted the preceding December, and to hear

objections from "interested persons . . . invited to be present

32
and be heard . . . " The criticisms heard fell into three

categories, according to the panel: (1) inaccuracies and

omissions, (2) defects in suggested means of implementing

9.29
certain panel recommendations, and (3) errors in findings
and recommendations related to particular computer projects.
The Committee took due note of all evidence, accepted the
panel's report, and discharged the panel, thereby putting
it out of existence. 33

In institutional affairs, the power to destroy implies
the power to create, and the following May the chairman of
the Research and Development Board "invited the members of
the ... Panel to act as consultants" in order to modify the
report they had given the Basic Physical Sciences committee
and produce thereby a final report which the Research and
Development Board might consider. Responsively, Fink,
Nyquist, and Trichel spent May 11th and 12th together in
Washington. "Acting in their capacity as consultants" to
the RDB Chariman, they made "minor revisions ... to meet
pertinent objection," eliminated certain criticisms for which
corrective action in the intervening nine months had been
taken, corrected various errors that had been brought to
light, and heard verbal reports from project representatives
in order to bring their information up to date.

The May sessions convinced the former panel members
that "no serious criticisms" of their earlier report had
been found, that original objectives had been sound, and
that the "broad purpose" to which their study was devoted
remained unchanged. ^^ The broad purpose had been to survey

9.30
existing computer projects being sponsored by agencies
in the Department of Defense and see what sort of integrated
and coordinated program should be developed to meet "the
over-all needs and objectives of all three Services ..."
As the consultants saw it, limitations must be placed on the
"laissez-faire" practices of the recent past and the pre-
sent exercised by different federal agencies. The appro-
priate organization to determine the limitations required
should be, the consultants felt, the Research and Develop-
ment Board.

One significant cause of the difficulties and counter-
criticisms they encountered in shaping their report is to
be found in the research and development philosophy the
panel embraced, as is revealed by changes made in the later
form of the report as a consequence of the passage of time.
Their philosophy was as different from Forrester's and
Everett's as was that of the Mathematics Branch and ONR.

A second cause of the problems they dealt with was
the sharpness of the counterattacks the first report re-
ceived from industry, government, and university repre-
sentatives. Events showed that while these RDB reprsenta-
tives of the scientific and industrial establishment were
seeking to curb and coordinate computer funding and pro-
gramming practices , they were unable to keep their own
houses in order.

9.31
Thus, a specific recommendation in the preliminary
report against building and leaving with the builder a
computer over which the Department of Defense retained

no rights of priority, title, or recapture was excised

35
from the final report, although the information that

the IAS machine would "remain at the Institute and be its

property with the Department of Defense holding no title

to the machine or to its use," appeared elsewhere in both

36
versions of the report.

Again, the panel had recommended that "a study pro-
ject be initiated on the subject of standardization of
input and output language of future machines." Apparently
this suggestion was premature and ill-advised, perhaps

considered impractical at the time by many experts, for

37
it, too, was deleted from the final report.

Similarly, the panel's objections to "a program for
copying a non-existent machine or a machine with an in-
definite completion date" and to a program "without a well

defined objective of value to the Departmett of Defense"

38
were stricken from the final report.

Hindsight permits perhaps too facile criticism to be

made of the Ad Hoc Panel's efforts to provide the Research

and Development Board with information and insights that

it might use to bring the progress of computer research

and development under more efficient, orderly, and prudent

9.32
control. History shows that these efforts failed, and it
shows also how difficult it was, even for accredited ex-
perts, to appraise the operations then going forward in
the newborn computer technology.

Dr. Lyman R. Fink had received his Ph.D. in electri-
cal engineering at the University of California (Berkeley)
in 1937 and was in the employ of the General Electric
Company at the time. Dr. Gervaise W. Trichel had obtained
an M.S. degree in engineering at MIT, a Ph.D. in electrical
engineering at the University of California in 1938, and was
then serving as a staff assistant to the General Manager,
Chrysler Corporation. Dr. Harry Nyquist had taken his Ph.D.
in physics at Yale in 1917, had worked in Vannevar Bush's
Office of Scientific Research and Development during World
War II, and had been associated with the Bell Telephone Lab-
oratories since 1934.

Yet, neither these men nor the Research and Development
Board nor the Secretary of Defense himself possessed the
leverage to effect the sort of coordination and control the
panel and presumably its parent Board sought. The cause
of this ineffectiveness lay neither in the newness of the
computer nor in the novel forms the emerging computer art
was assuming nor in any still unappreciated, still un-
dreamt developments in its future. It lay instead in the
deeply ingrained habits of carrying on highly technical

9.33

industrial and engineering affairs in the American economy. According
to these habits, the freedom of industrial and engineering enterprises to
enter into research and development projects with interested federal

agencies was a traditional and respected practice the over-
turn of which could only damage the dynamic mechanism that
enabled established interests to share in and contribute
to the wealth being distributed. If the coordinative man-
agement practices the panel recommended were to be follow-
ed, that wealth would be distributed in novel, arbitrary,
and capricious ways, restricting the freedom of action of
business, of universities .and of individual government
agencies. Such a consequence was as unthinkable as it was
insufferable in its implications of the extension of govern-
ment control; the major recommendations of the panel were
not followed.

Aside from its lack of political leverage in attempt-
ing to streamline the conduct of research and development
where computers were concerned, the panel also ran into
trouble when it attempted to evaluate the technical com-
petence of individual projects and ascertain whether they
were carrying on research and development operations in an
appropriate manner. Instead of consistently taking the
true measure of their worth — a task more approachable in
the light of hindsight than of foresight, — the panel
entangled itself in inconsistencies of judgment of which it
was not always aware. To conclude therefore that the panel

9.34
members, as well as their superior, the Committee on
Basic Physical Sciences, which accepted their report,
were (to put it bluntly) stupid is to fail utterly to
recognize the fundamental problem confronting all those
engaged in computer research and development. The prob-
lem was — and is — that of entertaining policy judgments
about the conduct of research and development affairs that
are sufficiently pertinent to the practices being carried
on as to provide effective control of those practices.

On the one hand were the policy ideals; on the other
hand were the daily, monthly, annual operations. The com-
parability and correspondence of the one to the other con-
stituted the truly formidable problems to which Forrester
and Everett were so sensitive, to which Perry Crawford was
so sensitive, to which Luis de Florez and Nat Sage and
Gordon Brown were so sensitive, to which Mina Rees and
Warren Weaver and John von Neuman were so sensitive, to
which Compton of MIT and Solberg of ONR were so sensitive.
Such problems, of course, are not restricted to computers
but permeate all scientific technology. In the case study
here at hand, Project Whirlwind, there was, as has been
shown, conspicuous divergence of opinion regarding the
proper correspondence that should be maintained between
policy and practice, and Forrester's opinions were

9.35
sufficiently extreme to cause ONR continuing trouble
because of the more orthodox views its programmers
held.

The Ad Hoc Panel's struggles to co-estimate policy
and practice are significant because they were so typi-
cal of the resistances that Project Whirlwind encounter-
ed when contemporary experts, handicapped by the:.fact
that the expert is always also the measure of contempor-
ary ignorance, sought to appraise its unorthodox pro-
cedures. For example, the panel criticised Project
Whirlwind in its earlier report for being without a
specific, practical end use. It was frank enough to
admit it was unable to find such a use, and said so
with such blunt candor that a reader might well infer
there really was none, and that the fault here lay princi-
pally with the machine and with the conduct of the project.

At the same time, the panel in both versions of the
report criticised the general conduct of computer research
and development because it did "not include sufficient
emphasis on real-time computation." ° Presumably Whirl-
wind did not qualify; "we may remark, in passing," said
the panel in its first report and in its final report,

"that there are at present no real-time electronic digital

41
computers in operation." And they were right; there

were no such machines in full operation. Was the panel

supposed to be prescient enough to realize that Whirlwind

9.36
soon would qualify? In 1949 they found Whirlwind I,
"as presently envisioned, ... a very large machine with
but five decimal digits capacity, extremely limited

memory capacity, and with as yet indefinite plans for

42
input and output equipment." Assuming these views to

be the ones the panel developed from its visit to the
Laboratory on May 26, 1949, they were sufficiently in
error and out of date a year later, in part as a con-
sequence of the progress of the Project, so that they
were replaced by more accurate and more optimistic in-
formation in the final report: "...somewhat limited in
its application due to its short word length (16 binary
digits) and limited internal storage (256 words) . Pro-
gress is being made toward 2048 word storage, however,
and certain applications for the machine have been set

forth in which the short word length is not a severe

43
handicap."

From the panel's point of view, the fact still re-
mained in 1950 that there were no real-time computers
in operation. True, "a large portion of the [MIT] com-
puter has been operating as a system since the fall of
1949," admitted the consultants, and "it is planned that
the machine will be available for useful computation in
the fall of 1950 . . . " 44 Yet following its survey, the

9.37
panel could conclude reasonably that these prospects
still did not invalidate their judgment that too little
effort was being devoted nationally to providing real-
time computers. The machine farthest along was the un-
orthodox, untried Whirlwind, but the panel did not appear
to find this very reassuring.

The panel knew that Whirlwind had indeed come far,
in one respect: it had found a mission since the Research
and Development Board had learned of its predicament in
1949. "Equipment currently is being connected," the con-
sultants were happy to point out in their confidential
final report, "for using the machine in real-time air

defense research for the Air Force, and plans for the

"45
fxrst year of machine use have been crystallized recently.

The striking feature of the Ad Hoc - Panel's predica-
ment in assessing the state of the art stands clearly
forth when one asks why the panel could not earlier find
a practical use for Whirlwind at the same time it was
bemoaning the lack of work in the real-time area of com-
puter applications. The answer would seem to be that the
panel could not free itself from the conventional attitudes
that prevailed. Instead, it pointed out sharply in 1949,
as has been remarked, that "the scale of effort on this
project is out of all proportion to the effort being ex-
pended on other projects having better specified objectives."

9.38
Considered years later, such conspicuous lack of
vision about the conduct of research and development
is readily seen, but the bases of this lack are often
times obscured. In this historical instance, the Ad
Hoc Panel's honest statements remove enough of the ob-
scurity to make it plain that their policy views were
not well adapted to describe or to cope with research
and development practices , yet appeared to reflect and
appraise the existing situation. In the final report
the panel softened its earlier criticism of the way in
which Project Whirlwind t,ad operated (and was continuing
to operate). Said the panel in June of 1950, after the
Air Force and Professor Valley's committee had come to
Forrester's timely rescue, "the technical direction of
WHIRLWIND seems to have suffered seriously by frequent
changes in objective and transfer of the project from
one division of the Naval Establishment to another. Al-
though the machine now has a definite objective, which
seems to us suitable and of enough importance to justify
the scale of expenditure contemplated, the current ob-
jectives have only recently been assigned. During much
of the time this project has been in existence, the scale
of effort has been out of proportion to that being ex-
pended on othltr projects having more definite objectives."^'
In accommodating its policy judgments to the changing

9.39
state of affairs, the panel took inconsistent positions:
there was not enough emphasis on real-time applications ,
there had not been enough emphasis in the past. Whirlwind
was, after all, only one such machine, and furthermore it
had been a machine without a task. As a machine without
a task, it could not merit the lavish research and develop-
ment support it had received, although now that it had a
task, it somehow did merit the support it was receiving.

Was the panel being realistic? Practical? Consider
the evidence in the realm of costs. The Raytheon "Hurricane"
prototype computer would cost an estimated $460,000, the
IAS computer $650,000, the Eckert-Mauchly UNIVAC $400,000
to $500,000, the National Bureau of Standards Interim Com-
puter $188,00, the U.C.L.A. Institute for Numerical Analysis
"Zephyr" Computer $170,000, the ONR-supported CALDIC com-
puter at the University of California $95,000, the famous
ENIAC $600,00 (three years old), its successor EDVAC, $470,000,
ORDVAC at the University of Illinois $250,000, and the Harvard
Mark III (just finished) $695,000.

The maximum or der-of -magnitude of cost appeared to be
a half-million to two-thirds of a million dollars. Whirl-
wind, however, would cost three million dollars, according
to current estimates and if all research costs were thrown
in since the beginning of the project, another three-quarters
of a million dollars should be added on top of that. To say
that Project Whirlwind was out of step is to put it mildly.

9.40
The fact that Whirlwind was costing as much as Forrester
had estimated two or three years earlier he had thought
it should, rounding off at about five million dollars
over a five or six year span, — whatever satisfaction
this correspondence gave Forrester was irrelevant in the
light of the criticisms of "excessive cost."

The policy judgment called for was clear: no com-
puter should cost so much, and all other research and
development experience with computers supported this
policy view. Whirlwind was way, way out of line. It
appeared that Forrester somehow had sucked MIT, and along
with MIT, ONR into carrying on research and development
at a fantastically inflated scale and cost. Again one
is reminded of the heated remark uttered years later by
a veteran engineer and project manager, "we are not go-
ing to have another Whirlwind!" Seen from this perspec-
tive, the criticisms of the Research and Development
Board's panel were temperate indeed. No wonder they had
urged in 1949 that if no practical end-use could be found
shortly, then the Project should be closed out!

Reflecting philosophically on these events a decade
and a half later, Mina Rees had remarked of the Project's
success in acquiring Air Force funding, "They were lucky."
They were lucky in the sense that for one set of reasons
Forrester and his associates had designed, developed, and

9.41
built a high speed, parallel type digital computer which
happened to be becoming operational , even if to a limit-
ed extent, at a time when, for a different set of reasons,
a serious national need had become evident. Without doubt,
the need of the Air Force had converged with the need of
Project Whirlwind at a time that was most opportune to both.
It is possible that other major uses could have been found
for Whirlwind, Forrester's search for applications had been
persistent and extensive and had become more intensive as
Whirlwind approached operational status. The curious his-
torical feature, nevertheless, is that dedication, persis-
tence, and industry — political as well as technical —
had produced a device at a time when there became apparent
a crisis in the national defense which it might help meet.
Project Whirlwind thus became a splendid example in support
of the novel and not-well-tested argument that research and
development, throughout its whole spectrum, should be sup-
ported for its own sake, because the use will always be found.
If this was a sound moral to be learned from the Whirlwind
experience, the lesson was not subsequently applied. In-
stead, the story of Project Whirlwind became submerged in the
larger drama of Lincoln Laboratory's crash programs, and
basic policy judgments about the conduct of research and dev-
elopment on a national scale remained as before.

FOOTNOTES CHAPTER

■'-Ltr., J. W. Forrester to J. B. Pearson, Deputy Director,
ONR f Devember 2, 1949; Memorandum, H. R. Boyd to J. W. Forres-
ter, sub j . : "Budget Basis," December 15, 1949.

o

*Ltr., R. J. Bergemann, to Head, computer Branch, ONR,
sub j . : "Project Whirlwind Progress: Comments On," December22, 1949.

3 Memorandum, C. H. Doersam, Jr., Synthetic Warfare Section,
SDS subj.: "Contract N7 ONR-38902, Project Hurricane, recom-
mendation of termination of, "November 1, 1950: Report, C.H.
Doersam, Jr., to code 920, subj.: "Trip Report to Raytheon
Manufacturing Company on 8 November 1950, "Project Hurricane,
C. H. Doersam, Jr." November 20, 1950: Internal Note, Samuel
Nooger, Head, Engineering Branch, SDC, N.D. [circa November 21, 1950

4
Memorandum, C. V. L. Smith to Code 434, subj.: "Letter of

J. W. Forrester, dated 2 December 1949, on the future financing

of digit 1 computer work at M.I.T.," January 16, 1950.

5
R. J. Bergmann, "Record of Visits January 1950-June 1950,"

January 12, 1950: C. V. L. Smith "Summary of conference on Trip,"

January 12 & 13, 1950.

Ltr., E. R. Piroe, Deputy for Natural Sciences, ONR, to
J. A. Stratton, Provost, Mit, 2 Feb. 1950? ltr., C. V. L. Smith
to J. W. Forrester, 2 March 1950: ltr., J. W. Forrester to J. A.
Stratton, 3 March 1950.

7
Interview, J. W. Forrester by the authors, October 26, 1967.

g

Ltr., Mina Rees to Kent C. Redmond, July 15, 1964.

q

C. V. L. Smith to George R. Stibitz, December 13, 194 :

ltr., Mina Rees to Professor Harold Hazen, MIT, December 28,
1948: ltr., H. L. Hazen to Mina Rees, January 10, 1949: ltr.,
H. L. Hazen to Mina Rees, January 12, 1949: J. W. Forrester to
Mina Rees, January 17, 1949: Mina Rees to Joseph E. Desch,
January 31, 1949: "Memorandum of Conversation with Dr. Alan
Waterman," March 7, 1949.

Ltr., C. V. L. Smith to J. W. Forrester, August 19, 1949.

J. W. Forrester to C. V. L. Smith, August 26, 1949.

12

Research and Development Board, Committee on Basic Physical

Sciences, "Report of the Ad Hoc Panel on Electronic Digital Com-
puters," PS 13/5, December 1, 1949. (Hereafter referred to as
the preliminary report.)

13

J. W. Forrester, "Comments on the Report of the Ad Hoc

Panel on Electronic Digital Computers of the RDB Committee
on Basic Physical Sciences," L-17, Servomechanisms Labora-
tory, MIT, January 13, 1950.

14

J. W. Forrester, "Information on Whirlwind I as re-
quested by Lt. Cmdr. Rubel, Research and Development Board,
in letter dated 7 February 1949, "Memorandum L-ll, Servo-
mechanisms Laboratory, MIT, February 15, 1949; J. W. Forrester
"Notes for RDB Committee meeting," August 29, 1949; in add-
ition, Memoranda L-3 and L-12, and Air Traffic Summary Report
No. 1 were made available to the panel.

15

J. W. Forrester, "Discussion of the Comments on Project

Whirlwind made by the Ad Hoc Panel on Electronic Digital
Computers of the Basic Physical Science Committee of the Re-
search and Development Board," Memoranda L-16, Servomechisms
Laboratory, MIT, January 13, 1950.

Memoranda, Code 434 to Code 102, signed by A. E. Smith,
Acting Head, Computer Branch, ONR, subj . : "Confidential Re-
port of the Ad Hoc Panel on Electronic Digital Computers,
RDB, Comments on," 20 December 1949.

17 .

Diary Note, A. T. Waterman, subj.: "Whirlwind Conference

held at Massachusetts Institute of Technology," March 6, 1950:

C. M. [Carl Muckenhaupt] , "Record of Visits January- June , 1950,"

March 6, 1950.

18

Diary Note, A. T. Waterman, subj.: "Whirlwind Conference

held at Massachusetts Institute of Technology," March 6, 1950:

Memorandum, C. V. L. Smith, "Summary of Conference on Trip,"

March 6-7, 1950.

19

Memorandum, C. V. L. Smith, "Summary on Conference on Trip,"

March 6-7, 1950.

20

Memorandum, N. McL. Sage, April 7, 1950.

21

Ltr., N. McL. Sage to Chief of Naval Research, May 5, 1950;

Memorandum D-2 3, Jay W. Forrester to Head, Computer Branch,

Mathematical Science Division, ONR, subj.: "Request for Funds

for Contract N5ori-60 for the Period July 1, 1950 through June

30, 1951," May 3, 1950.

22

Ltr., C. V. L. Smith to N. McL. Sage, June 26, 1950.

23

Ltr., J. W. Forrester to N. McL. Sage, July 6, 1950;

N. McL. Sage to J.W. Forrester, July 11, 1950.

24 Memorandum for File 6345 by Paul V. Cusick, DIC, MIT,
October 6, 1950; Amendment #12, T. 0. #1, Contract N5ori-60,
November 14, 1950.

25 Ltr., N. McL. Sage to C. O., ONR, February 23, 1951; inter-
view, J. W. Forrester and R.R. Everett by the authors, October 26,
1967.

26 Ltr., J. W. Forrester to C. O., ONR, September 25, 1951.

27 Amendment #13, T. 0. #1, Contract N5ori-60, June 28, 1951;
Amendment #14, T. 0. #1, Contract N5ori-60, June 28, 1951.

28 Amendment #16, T. O. #1, Contract N5ori-60, 26 March 1952;
Amendment #19, T. 0. #1, Contract N5ori-60, 18 April 1953.

^Interview, J. W. Forrester by the authors, October 26, 1967.

JU Research and Development Board, "Report on Electronic
Digital Computers by the Consultants to the Chairman of the Re-
search and Development Board, PS 13/8, June 15, 1950.

31 J. W. Forrester, "Statement of Status of Project Whirl-
wind Prepared for the Research and Development Board," L-24,
Servomechanisms Laboratory, MIT, May 10, 1950.

32 "Report on Electronic Digital Computers by the Consultants
to the Chairman of the Research and Development Board," June 15,
1950, p. vi. (Hereafter referred to as the final report.)

J Ibid. , pp. vi-vii.

34 Ibid., p. viii.

Cf. pp. 9 and 9 respt. , of the preliminary (December 1, 1949)
and final (June 15, 1950) versions of the report.

3 Cf. pp. 44 and 45, respt., of the preliminary and final
versions of the report.

37 Cf . pp. 8 and 23 of the preliminary report with pp. 9 and
22, respt., of the final report.

Cf . pp. 9 and 9 respt. , of the preliminary report and the
final report.

39 Cf. pp. 11, 39 in the preliminary report, December 1, 1949.

Cf_. pp. 34 and 33 respt. , in the preliminary report and the
final report.

Cf. pp. 15 and 13, respt., in the preliminary report and
the final report.

Preliminary report, p. 53.

43

Final report, p. 54.

44

Ibid.

45
Ibid.

46

Preliminary report, p. 39.

Final report, p. 40.

Chapter Ten

ADSEC AND WHIRLWIND

The Air Force phase — the "Valley Committee,"
Project Lincoln, and SAGE — was the beginning of the
triumphant end for Project Whirlwind, for its greatest
successes and the vindication of Forrester's and
Everett's research and development policies occurred
during this period. None of the principals was in
a position to perceive in 1950 that the end was com-
ing into sight, of course, even though the computer
had found a specific mission, nor did they realize
that it would take the form of a subtle transformation
of the Project as a consequence of gradual alteration
of the R&D climate in which the Project originally
had achieved its identity and begun to flourish. This
transformation of the Project which occurred as it
adapted to the changing R&D climate and began to
live with its own successes produced at the same time
an assurance and a sophisticated maturity of operation,
and intimations that the frontier in which the Project
had come to life was passing on and that a more
settled way of life was approaching, even while the
Project continued to busy itself with the technical
innovations which were its prime concern. Several
years were to pass before Forrester himself became
convinced that the settled way of life which loomed
was not for him, and it was not until 1956 that he
set out again, to find the frontier.

That the Air Force should have become involved
in 1950 in determining Project Whirlwind's ultimate

10.1

10.2

fate, rather than the Navy, was the circumstance of
a current of events generated quite independently
from those that had brought Whirlwind into existence.
These events are worth looking at because they
created the conditions in which Whirlwind was to
succeed, because they were not anticipated by the
Navy and the MIT managers who were bringing Whirlwind
along, and because they were central in setting in
motion the changes in the R&D climate which were
subsequently to transform the Project.

In one sense, although not a particularly profound
one, the antecedent events began to take form when war
in the air became a practical, prospective possibility
after the Wright brothers ushered in the age of heavier-
than-air flight at Kitty Hawk, North Carolina, on
December 17, 1903. But it was not until the Second
World war that the technology of aerial warfare be-
came sufficiently advanced to pose the threat of aerial
attack upon the United States by foreign, land-based
bombers. Consequently, the American people, their
government and military leaders, sustained by a
confidence derived from geographical isolation and
continental dominance, had not during the years preceding
the War seen any vital need to plan, research, and build
an integrated continental air defense system.

In the immediate post-war days, the American
people, basking in the warmth and fellowship of the
Allied victory and the birth of the United Nations
and residing in a homeland unscathed by the War, did
not foresee the dangers inherent within the changed
international balance of power. Increasing difficulties
with Russia, however, coupled with the advances made
in military technology during V/orld War II, aroused
Americans from the euphoria of victory and compelled
them to give serious consideration to the nation's
vulnerability to attack from the air in the new age
of long-range bombers, missiles, and atomic warheads.

10.3

This growing awareness was reflected by the
United States Air Force in December of 194-7 when
General Hoyt S. Vandenberg, Vice Cnief of Staff,
in a letter to Dr. Vannevar Bush, Chairman of the
Research and Development Board, expressed anxiety
over the lack of an adequate national air defense
system. The Vice Chief of btaff discussed a projected
plan which, it was estimated, by using obsolescent
radar of world War II design would give the nation
by the end of 1952 some degree of increased protec-
tion, pending modernization with post-war developed
and manufactured equipment. It was this last, how-
ever, which seriously worried General Vandenberg,
for he was of the belief that the existing state of
electronics research and development in air defense
and guided missiles would not produce advanced designs
available for production until after 1953 • This was
particularly critical, since the mid-fifties were

estimated to be a period of especial danger for the.

p
nation. With these fears in mind, Vandenberg put

three questions to Dr. Bush: Were new developments
coming along at the best rate possible? vvas the
Army-Navy-Air Force program properly balanced? Were
there serious technical deficiencies in the military
programs which required immediate correction?-^

Reacting to the seriousness of the problems
raised by the General, Dr. Bush asked the "Subpanel
on Early warning" of the "Radar Panel" of the "Commit-
tee on Electronics" of the "Research and Development
Board" to investigate the existing state of national
air defense and to prepare a careful analysis of the
system or systems-in-being and contemporary programs
seeking improvement in the national air defense pos-
ture. The Subpanel concluded that existing research
and development programs were ample, but insufficient

10.4

funding and difficulties caused by the two-year fiscal
restrictions imposed by the Constitution on military
appropriations made impossible maximum progress in
the implementation of long range plans and programs.
An improved air defense system could be made avail-
able even earlier than tae Vice Chief of Staff had
predicted, the Subpanel opined, provided existing
research and development programs were given immedi-
ate emphasis and accelerated by increased funding.

h.

Otnerwise, a new system would still be years off.

The existing system was inadequate and decentral-
ized. It comprised separate defense areas, each with
its own radar and each responsible for locating,
identifying, and intercepting hostile aircraft that
had penetrated the area. The radar was automatic,
but accumulation, analysis, and interpretation of the
data received, whether from radar or other sources,
were performed by men; hence tae defense of any single
area was dependent upon the speed, competence, and
efficiency of the forces responsible for it, and these
were not unlimited. -^

'weaknesses within the system were numerous. The
radar could easily be saturated. There were communi-
cation difficulties between machines and operators.
Serious gaps existed in lo.w-altitude coverage, and
the employment of smaller radars to fill such gaps
was not considered feasible, since additional data
would impose a heavier burden upon already overtaxed
control centers. Voice-radio communications between
stations and control centers were not reliable. Along
the Eastern and Western sea frontiers, arrangements
for early warning of approaching aircraft were inade-
quate. The primary weakness, nowever, was the limited
ability of the system to process, organize, and use
the information it gathered.

10.5

Air Force interest in the condition of the
nation's air defense system was boosted in March of
194-8 when the Joint Chiefs of Staff gave it primary
responsibility for defense against air attack. This
assignment of primary responsibility, coupled with
the Air force's definition of air defense as "all
measures designed to nullify or reduce the effective-
ness of the attack of hostile aircraft or guided
missiles" after they had become airborne, left the
Air Force with almost exclusive responsibility for
continental air defense. The only exception, anti-
aircraft artillery, continued a functional assignment
of the United states Army.'

In September of 194-9* Americans heard the alarm-
ing news that Soviet Russia had successfully detonated
an atomic bomb the preceding month. This broke the
American monopiy several years earlier than had
been anticipated, shortly thereafter, intelligence
sources suggested that Russia possessed sufficient
air carriers to penetrate existing American air defenses,
and might possibly within the immediate future match
the United States in both the number and size of nuclear
weapons and the quality and quantity of jet bombers.
For informed Americans this increased Russian mili-
tary power in the midst of Cold War tensions trans-
formed a threat into a clear and present danger and
compelled the Air Force to reappraise the state of
the nation's air defenses. The reappraisal was
carried out by several study groups, each of which
analyzed the existing systems and equipment with a
view to determining their weaknesses and recommend-
ing corrective measures. Nor was the Air Force alone
in its fears, the Department of Defense, reflecting
similar apprehensions, also established a study
group under the aegis of the Weapon Systems Evaluation

10.6

Group. A new sense of urgency regarding air defense
was spreading throughout the federal military estab-
lishment .

Two ad hoc programs were mounted by the Air
Force. One, conducted under contract by the Bell
telephone Laboratories, was concerned with possible
immediate improvements of the existing system and its

Q

equipment. The other studied possible new approaches
and new systems. It came into being as the result of
actions generated simultaneously in the military and
in the scientific communities. General Vandenberg
was the military proponent and Professor George E.
Valley of the hassachusetts Institute of Technology
v/as the scientific proponent.

As Air Force Chief of staff, General Vandenberg
became increasingly apprehensive after the Russian
atomic success of August, and he urged upon his
colleagues the desperate need to act immediately to
develop and construct a continental air defense
system that could cope with the offensive potential
of get bombers and missiles. The Vice Chief of
Staff, General huir b. Fairchild, relayed these fears
to a civilian consultative body, the Air Force Scien-
tific Advisory Board (SAB), and added his cuperior's
request that the Board undertake "a continuous study
of the technical aspects of the air defense of the
Continental United States."-^

Concurrently, Irofessor Valley, a pnysicist and
a member, of the bAB, was in the latter capacity •
involved as both observer and participant in test-
ing and discussing the existing national defense
system. Disturbed by the confusion and poor under-
standing of the problem that appeared to him- to exist,
Valley proposed in November, 1948, that the Board
create an "Air Defense Committee" to conduct a tech-

10.7

nical investigation of the existing system and to
determine "the best solution of the problem of Air
Defense." 10

In response to Vandenberg's and Valley's urging,
the Air Defense System Engineering Committee (ADoEC)
came into existence in order to provide a mechanism
for examining the apprehensions of military and sci-
entific planners, assessing the vulnerability of the
nation to aerial attack and proposing appropriate
solutions. The chairman of ADSEO was Professor
Valley, whence the frequent informal references to
ADSEC as the "Valley Committee." Its members included
George Comstock, Allen F. Donovan, Charles S. Draper,
Henry G. Houghton, John Marchetti, and H. Guyford
St ever.

All of the Committee's members save two — George
Comstock and John harchetti — were members of the SAB
and all came from the northeast section of the nation.
The geographical concentration was deliberate, intended
to permit easy consultation and recourse to the facili-
ties of the Air hateriel Command and the Continental
Air Command at the Air Force Cambridge Research
Laboratories in Cambridge, Massac ausetts. Valley
had recommended ohis arrangement on the assumptions
that the members of the committee would be able to
serve only on a part-time basis, that solution of
the problem would be difficult and long, and that
experiments which would require the use of radar and
aircraft would be conducted. Moreover, he thought,
conditions on the West Coast might be sufficiently
different to require a special investigatory group
for that region.

Events following the creation of ADSEC proceeded
on at least two levels significant to the story of

10.8

Project Whirlwind: one was the level of long-range,
institutional policy development; the other was the
level of unfolding scientific, technical, and mili-
tary insights into the practices that might be pur-
sued to achieve a working defense system which would
constitute a defensive force-in-being. Against aerial
attack over the North Pole no standing army, no navy
at battle stations could provide the needed force-
in-being. Aerial attackers would have to be detected
in the air and defeated in the air before they released
the bombs and rockets they carried. The first problem
thus became one of scientific technology, even before
appropriate military and naval operations could be
instituted. ADSEC indeed had its work cut out for it,
nor was there sure warrant in advance that ADb.bC or
any other responsible agency could effect a practical
solution before time ran out.

Before further developments on this scientific
and technological level of events are pursued, a
glance at major developments on the institutional
policy level uuring the early 1950' s, following the
creation of ADSLC, may provide a longer-range per-
spective and a useful framev.ork in which to examine
the circumstances that shortly swept up Project
Whirlwind. Whirlwind became involved in these affairs
early in 1950 and subsequently became an operation so
wholly transformed that even to its leader it changed
its original identity as a research-and development
enterprise before the job was done. Both the trans-
formation and trie consequences were profound, and the
policy-level events responsible, following the forma-
tion of ADSEC, occurred in one-two order. First, they
took the form of the independent researches conducted
by ADSEC and by the committee operating under the aegis
of the Weapon Systems Evaluation Group. Second, they
took the form of the conclusions these two bodies reached.

10.9

Their conclusions reinforced each other and con-
firmed Air Force fears about the inadequacy of the
nation's air defenses. In extremely strong language,
the SAB subcommittee compared the existing system "to
an animal that was at once 'lame, purblind, and idiot-
like, '" insisting that "of these comparatives, idiotic
is the strongest. It makes little sense for us to

strengthen the muscles if there is no brain; and given

] 2
a brain, it needs good eyesight." Translated into

technical language, this meant that an adequate air
defense system needed not just improved interceptors,
ground-to-air missiles, and antiaircraft artillery,
but improved radar and advanced command-and control
centers.

The cumulative impact of the two studies and their
separate but reinforcing conclusions produced a request
from the Air Force in December of 1950 to the hassa-
chusetts Institute of Technology, asking the Insti-
tute to create a laboratory which would dedicate
itself exclusively to research and development pointed
to the design and construction of an adequate national
air defense system. Early in the next month, the SAB
added its voice to the Air Force request and in addi-
tion asked the Institute to undertake a more inten-
sive study of the technical problems connected with
continental air defense. From this latter request

evolved i-roject Charles, an investigation conducted

13
between February and August of 1951

Upon the approval of its board of governors,

the Institute acceded to the request of the Air Force

and established Project Lincoln, subsequently better

known as Lincoln Laboratory. Ihysical facilities to

house the Laboratory were suortly constructed at

Hanscom Air Force Base in the nearby Bedford-Lexington

region. The Laboratory's immediate task was to imple-

10.10

merit trie defense concepts formulated by the Air
Defense Systems Engineering Committee and Project
Cnarles. This immediate task plus its ov/n researches
led, in the months and years that followed, to the
Lincoln Transition System and then to the semiauto-
matic Ground Environment (SAGE) system, a total
continental air defense system which was at once "a
real-time control system, a real-time communication
system, and a real-time management system," using

"digital comouting systems to process nation-wide

14-
air-defense data."

Project Whirlwind — to return to the technological
level of events — had become involved with AD'-^C immed-
iately after the formation of the committee, terry
Crawford, visiting iroject Whirlwind on January 20,

1950, had informed Jay Forrester of the creation of

15
the committee. Within a week there occurred a

coincidental chance meeting on the HT campus of

two of its faculty members, Dr. Valley and Dr. Jerome

16
B. Wiesner. As Valley recalled it years later,

Wiesner a ked him conversationally how things were
going, one remark led to another, and soon he was
indicating to Wiesner his need for an information-
gathering and information-correlating center that could
organize great numbers of diverse pieces of information
with extreme rapidity. Wiesner suggested he take a
look at Jay Forrester's operation to see if it would
have anything of value to offer. Valley followed up
this lead and found a machine so promising that he
and his fellow ADSEC members seriously investigated
the prospects and decided to support a test narness-
ing of 'whirlwind in order to determine whether geo-
graphical information received by radar scouting
stations could be transformed into tactical information
and directions that would enable a fighter to inter-

10.11

cept a bomber long before the latter reached its
■car get.

For Forrester, the involvement of Whirlwind in
ADbEC affairs came to pass within a few days after
Crawford told him the committee had been formed. On
Friday, January 2'/, he was having lunch with Frofes-
sor v.iesner. At the lunch table, as Forrester wrote
in his notebook later, with customary attention to
recording developing events, "we were joined by
George Valley to discuss his committee work on Air
Defense bystem Engineering." ' As yet, Forrester
had no special reason to believe that decisions of
particular consequence would follow. Tnis was but
another of several "leads" he was pursuing as part
of his probing operation to find a use for Whirlwind
and incidentally either relieve or justify (or both)
his heavy dependence on O&R for funds.

Wiesner's role appears to have been that of the
unobtrusive broker wno, having been deliberately
instrumental in setting events in operation, fades
gracefully into the background and equally deliber-
ately passes on to the principals involved the
responsibility of carrying; affairs forward. During
the course of their lunch, Valley explained the pur-
pose of his committee, outlined the weaknesses of
the existing air defense and eai\Ly warning system,
and discussed "his plans for a research project
involving the Cambridge Field btati n, Draper's
group, and possibly the Dipatal Computer Laboratory i"
After lunch, he accompanied Forrester to the labor-
atory. There, after observing whirlwind 1 operating
with test storage, the two men discussed the possi-
bility of the computer project's participation in
the work of the Valley Committee. Valley upon this
occasion expressed his intention to pursue the matter

10.12

further, Forrester gave him copies of the L-Reports,
L-l and L-2, that he and Everett had written more
than two years earlier, in the early autumn of 194-7 ,
detailing how computers might be employed to handle
interception problems in antisubmarine warfare.

They arranged to get together the following week
with certain of Valley's defense-work associates,
and Valley spoke of using this situation to help him
prepare the inevitable proposal to Washington. "He
seemed quite interested in possible use of wnirlwind
I for analyzing data from a chain of doppler radar
stations which would give range rate only," Forrester
noted at the time.

Three days later, on Monday, January $0, Valley
returned to the laboratory, bringing with him three
other members of the committee, John Karchetti of
the Air Force Cambridge Research Laboratory, H. Guy ford
btever, and Charles S. Draper, both of the Institute's
aeronautical engineering faculty. 7 Accompanying the
group was Eugene Grant, a colleague of Marchetti but
not a member of the committee. The visitors reviewed
some of the material Forrester and Valley had dis-
cussed at their earlier meeting, observed the computer
and the "storage tube television display," and dis-
cussed the use of V.hirlwind I in some of the "trial
systems" the committee was considering.

The trend of the discussion revealed that Valley's
initial interest, which had been heightened by his
reading of the L-ixeports and had brought him back to
the laboratory, was shared by his colleagues on this
occasion. To Forrester, his visitors seemed "very
enthusiastic about the prospect," and tnis opinion
was given substance by the group's discussion of the
question of funding. "Valley and the other men," wrote
Forrester that day, "seemed well aware of the fact

10.13

that they would be called upon to share some of the
basic :;!>600,000 a year budget for the laboratory, plus
additional charges for special work on their own
project. ... The Committee is apparently meeting
in Washington in the next few days to crystallize the
matter further."

Forrester's observation was quite accurate.
ADbLC became committed almost immediately to the use
of Whirlwind I, as was evidenced by its meeting of
February 17, which was attended by Forrester,
Fahnestock, and Sverett. The discussion on this
occasion focused on the possibility of attempting
some trial interceptions, using the Bedford MEW radar
to track the aircraft involved and feed the data to
Whirlwind I to process. The committee's enthusiasm
was such as to induce Forrester to enter in his
record of the meeting, the cautionary observation

that "we shall have to be careful not to be driven

21
into a situation demanding more than we can deliver."

Of greater significance to Forrester and his

colleagues was the information Valley had given

Forrester just a few days before the meeting, lihen

informing Forrester that formal approval for ADbEC

to go ahead with its work was anticipated by March 1,

Valley had confidentially disclosed that the Air

Force, if required, might assume the entire cost

22
of keeping the project going.

As a consequence of this groundwork laid early

in 1950, Valley was in a position formally to offer

the dollar support that neither OKR's nor iuIT's top

management considered feasible to continue longer

to provide under the long-standing relationship that

had begun with de Florez' office and since passed

through many Kavy hands. The apparent withdrawal of

10.14-

a measure of hIT support from Forrester on this occa-
sion, as it may have appeared from OMi's vantage
point, was tempered by the awareness in otratton's
office of what Valley was doing in response to his own
feeling of concern, as well as that of various offices
in the Pentagon and other Executive Branch agencies,
regarding the unsolved problems of continental defense,
including temporary "quick fix" proposals. Project
v.nirlwind now would have an opportunity to prove
itself and relieve itself of the sort of onus cast
upon it by the I-iDB Ad Hoc Panel's preliminary report.

Forrester was able to go into the ftarch meetings
with ONR with reasonable assurance of Air Force finan-
cial support and, moreover, with the confidence that
at long last, he was to be given the opportunity to
demonstrate the concept he and Perry Crawford had
consistently advanced: the usefulness of the high
speed digital computer to a command-and-control center.

Actually, the ease with which Project Whirlwind
became incorporated into the program of the Valley
Committee was understandable and logical. In several
different ways, the Project was uniquely qualified
to serve the committee's needs. It was conveniently
located geographically. It had brought to the edge
of operational status a high speed digital computer
which was not only appropriate but even essential to
some of the tests the committee envisaged. It
possessed a pool of scientists and engineers trained
and experienced in digital-computer research and ■
development. Lastly, its long-standing commitment
to practical problems, wnich had caused warren Leaver
in February* 194-7 » to ask probingly whether it was
trying to produce "biscuits" or "oake," which had im-
pelled Forrester and v/verett several months later to
write those first L-Peports, on naval warfare, and

10.15

which later had involved the Project in investigations
for the Air Porce into air traffic control had given
it a unique expertise. In naval warfare problems and
in the "application of digital computers to the long-
term Common System of military and civil air traffic

control," lay api: lie at ions "in many ways similar to

23
air defense. ^ The problem common to all was the

processing and organizing of information to provide

"the ability to see a complex situation as a whole."

Possessing computation speeds immeasurably greater

than man's, the computer could "scan the component

pieces of information which make up the system so

rapidly that it is able to approximate a continuous

24
grasp of the whole situation." Therein lay its

value to naval and air defense as well as to air
traffic control, for these represented the obverse
and reverse of the same problem.

Of greater significance and importance to the
air-defense phase of Project Whirlwind was the recog-
nition gained from the air traffic control study that
the approach to such problems had to be systems-oriented
rather than component-oriented. When in March of 1949 »
Project Whirlwind had contracted with the Air Force
to undertake the air traffic control study, "there was
little or no background in the use of a digital com-
puter as the control element of a physical system
and very few people ... experienced in this sort
of work." The investigators were working in virgin
territory, and as they gained experience and understand-
ing, it became more and more evident that the appli-
cation of the digital computer to air traffic control
was "as much a systems problem as a computational
problem." Totally "new concepts of the whole system"
were required if maximum results were to be derived

10.16

from using the computer. ^ The insights into systems
engineering gained from the air traffic control study
were of great benefit to the engineers of Project
whirlwind who later, as a part of Lincoln Laboratory
were to contribute vitally to the development of the
SAGE air defense system.

Ltr., Gen. Hoyt S. Vandenberg, VCS, bbAF,
to Dr. Vannevar Bush, Chairman, RDB, December 9, 194-7.

p

Memorandum K-1810, A. P.Kromer, subj.:
"Mnutes of Joint hIT - IBM Conference, Held at Hartford,
Connecticut, January 20, 1953," January 26, 1953.

^ Ltr., Gen. Hoyt S. Vandenberg, VCb, bbAF, to
Dr. Vannevar Bush, Chairman, RDB, December 9, 194-7.

4

Ltr., W. L. Barrow, Chairman, Ppnel on Radar,

Committee on r..lectx , onics, HDB, to iMorman L. winter,

Executive Director, Committee on Electronics, HDB,

January 5, 1948.

5

History of the Air Force Cambridge Research
Center , 1 July - 51 December, 1953, Vol. XIX, part 1,
pp. 247-252.

La Verne E. 'Woods, "The Lincoln System," ADC
Communications and Electronics Digest , IV (January,
1954) pp. 4-11 , "cited ibid ., pp. 249-2 50 .

7

C. L. Grant, The Development of Continental
Air Defense to 1 September 1954 , UoAF Historical
studies! Ho. 12"b \ pp. 14-18.

Operational I Ian , Semiautomatic Ground Environ -
ment oystem for Air Defense , HEDADC, harch 7i 1955 i P» v .

9 hemo for DCS/C, DCS/P, DCS/0, and DCS/PI, HEDUSAF
from General fruir S. Fairchild, VCS, L3AP, subj.: "Air
Defense Tecnnical Committee of the Scientific Advisory
Board," December 15, 1949, cited in History of Air
Force Cambridge Research Center , Vol. XV, Part 1,
Appendix 11.

Ltr., G. E. Valley to Dr. Theodore Von Karman,
November 8, 1949.

11

Ibid.; ltr., General huir d. Fairchild to Com-
manding General, Air hateriel Command, subj.: "Air •
Defense System Engineering Committee, Scientific
Advisory Board to the Chief of Staff, l».S. Air Force,"
January 27, 1950.

x ADSEC Report, "Air Defense System," October
24, 1950

15 Ltr., General Hoyt S. Vandenberg, COFS HEDUSAF
to Dr. James R. Killian, President, hlT, January 19, 1951.

14 R. R. Everett, C. A. Zraket, & H. D. Benington,
"Sage - A Data Processing System for Air Defense," reprint
from Proceedings of the Eastern Joint Computer Confer -
ence , Washington,' D. C. , December, 1957* p. 148.

15
■^ J. W. Forrester, Computation Book No . 49, p. 77

entry for January 20, 1950.

Interview, Professor George E. Valley by the
authors.

17
' J. W. Forrester, Computation Book Mo . 49 , p. 83,

entry for January 27, 1950.

18 Ibid .

19 . ■
J Dr. Stever was at that time an associate pro-
fessor of aeronautical engineering and a member of

the licsAF Scientific Advisory Board. Dr. Draper was
director of the BIT Instrumentation Laboratory and
in 1951 became aead of the Department of Aeronautics.

20

J. W. Forrester, Computation Book No . 49 , p. 84,

entry for January $0, 1950.

21

J, to. Forrester, Computation Book No . 49 , p. 88,

entry for February 21, 1950.

22

J. to. Forrester, Computation Book No . 49 , p. 86,

entry for February 15, 1950.

25
-^ C. R. V/ieser, in Lincoln Laboratory Quarterly

r'rogress Report , Division 6 — Digital Computer , June 1,

1952, pp. 6-9.

24

C. R. toieser, "Digital Computers in Control

bystem," Report R-181, oervomechanisms Laboratory,

KET, April 27, 1950.

25

^ C. R. toieser, in Lincoln .Laboratory Quarterly

Progress Report , Division 6 — Digital Computer , June 1,
1952, pp. 6-9. ■

Chapter Eleven

INTERNAL STORAGE PROBLEMS

When ADSEC had come on the scene during the winter of 1949-50,
the engineers and technicians in Project Whirlwind were involved in
fabrication, assembly, and testing operations that kept them too
busy to draw any conclusions other than the pragmatic day-to-day
and week-to-week conclusions that genuine progress (as well as
apparent physical progress) was being made in all areas— with the
possible exception of the electrostatic storage tubes, and even the
latest designs of these were performing impressively. Forrester
found it entirely reasonable to assert a decade later that World
War II electronic-circuit development experience had left a fairly
straightforward prospect for computer design, but that satisfactory
internal-storage elements had posed an entirely different problem.
The mercury delay line, as well as the rotating magnetic drum
developed by Electronics Research Associates in St. Paul, were both
too slow for Whirlwind's real-time response needs dictated origi-
nally by the aircraft simulator and then held to by Forrester and
his associates in the knowledge that it was an essential feature of
the sort of general- pur pose machine they visualized. From 1947 to
1950, one could argue, they and Perry Crawford appeared to be about
the only ones who could see actual, attainable applications for
such a computer; yet this gave them no cause to depart from their
conviction. To others in the growing computer business, their

11.1

11.2

stubborn conviction was part of the irregularity their project
exhibited, part of the youthful and immature enthusiasm that
"tainted" them.

Their early analysis of "moving- target indicator" tube designs
generated out of the British invention of radar led them to con-
clude that this type of storage system, which initially looked
promising, suffered from limitations that became prohibitive when
examined in greater detail. This was the Williams storage tube,
in its basic concept, and when Forrester considered its prospects
for further development and refinement as a digital-computer ele-
ment, he concluded, rightly or wrongly, that it "inherently lacked
the high signal levels, the high signal-to-noise ratio, the ability
to give good signals from the noise, that we would require for our
high-reliability application . . . ." In consequence, recalled
Forrester, "we did not stay with the Williams tube idea for very
long . . . ." 2

In this state of the art of the key storage elements of the
computer, Forrester and his associates had early come to recognize
the "serious disadvantages" of any device, including the MIT
Radiation Laboratory tube, similar in some respects to the Williams
tube, that the Project finally settled upon for further development
work. Forrester's chariness from the start to state when Whirlwind
would incorporate reliable storage tubes indicates how serious he
felt the contemporary lack of a high-speed, reliable, internal-
storage element to be, with respect to its effect upon the relia-
bility and speed of operation of the entire computer.

11.3

In 1947 Forrester even had briefly considered using inter-
connected "gas glow discharge cells" in three-dimensional array,
for these offered such advantages, at least theoretically, as
"high initial breakdown voltage," "... low holding voltage, and
low forward impedence after breakdown . . . ." But investigations

were "soon dropped because of the variability of the individual

3
cells with time and from sample to sample."

Once Forrester had committed his project to electrostatic
storage tube research and development, he publicly buried his
earlier misgivings in order to get on with the specific course of
action that to him appeared to be least impractical so far as an
internal -storage device was concerned. His experience and instincts
in design continued to keep him apprehensive and sensitive, appar-
ently, to the complexion of research events that followed during
1947, 1948, and 1949. During this period, as the electrostatic
tubes began to be developed, new problems continually were emerging,
were being overcome, and were being replaced by other problems.
Forrester, and later Everett, closely watched the design progress
from the earliest research-demonstration tube forms, through ver-
sions in 1948 that Forrester later called "experimental research
tubes," to a more compact form of late 1948-early 1949, which
possessed a shorter beam-throw from the writing and reading elec-
tron "gun" and from the holding "gun" to the storage surface. The
first gun was "an ordinary type cathode-ray gun." The second
operated at a low voltage to furnish "a uniform flood of electrons

11.4

over the entire surface," thereby keeping the different charges on
the mosaic points, or "spots," of the storage surface from fading
away. These charges , depending on their value, were the "bits" of

alternative, binary-code "information" the tube was expected to

4
store, relinquish, or alter on demand. Access to a tube's infor-
mation (including obtaining a "read-out" of a single charge, or
bit,) should have been six microseconds, but ten to 25 seconds was
the performance of the tubes that were available early in 1949.

The laboratory work was proceeding forward perhaps about as
well as might be expected, but within Forrester's own thoughts,
out of sight and knowledge of his Project Staff, except for per-
haps Everett, was disquiet and a continuing watchfulness. The
electrostatic storage tube was analogous to a patient making
reasonable — or perhaps not quite reasonable — progress, yet by no
means out of danger of a relapse.

Forrester had early recognized that rapid access time must
accompany a vast storage capacity, hence he saw the advantages of
the geometry of arrangement suggested by three-dimensional coinci-
dent excitation storage, such as the gas glow discharge cells had
seemed at one time to offer. When, in the spring of 1949, he saw
an advertisement announcing the industrial availability of a magne-
tic material "Deltamax," there recurred to him the possibility of
reviving his three-dimensional, coincident-current system, employ-
ing reversibly magnetizable intersections. In June he began to
make entries in his computation book that indicate he was at work
in the lab again.

11.5

"Shortly before the Air Force came into the picture, from
where we sat," recalled a former graduate student in the Laboratory,
"Jay took a bunch of stuff and went off in a corner of the lab by
himself. Nobody knew what he was doing, and he showed no inclina-
tion to tell us. All we knew was that for about five or six months
or so, he was spending a lot of time off by himself working on
something. The first inkling I got was when he 'came out of
retirement' and put Bill Papian to work on his little metallic and
ceramic doughnuts. That was in the fall and winter of 1949-1950."

Forrester at the end of June, 1949, had begun to test rings,
or cores, of the Deltamax magnetic material he had seen advertised
by a subsidiary of the Allegheny Ludlum Company, the Arnold Engi-
neering Company.

He could see advantages in certain theoretical prospects*
The question was, would the magnetic materials behave every time
the way they should in principle, or, like the electrostatic tube,
would they prove delicate, operable at a peak (and essential) level
of performance for only a relatively brief time, and difficult to
produce and maintain at the quality and reliability levels required?
In principle, a magnetic core should be capable of holding either
of two electromagnetized states and should require sharp differences
of energy to change from one state to the other. This property,
describable in more technical terms as a rectangular hysteresis-
loop effect, would provide the binary "bits"--the "yes" and "no"
information — required. Such information could be tapped ("read")

11.6

or altered ("written"), depending on the kinds and strengths of
pulses fed to the magnetic core.

Everett recollected it was early in the summer of 1949 that
Forrester had brought the possibilities to his attention in the
form of a plan for a planar array of cores. Shortly, he showed
Everett a three-dimensional plan, and by early August he had satis-
fied himself that tiny cores an inch or less in diameter ought to
perform well if composed of suitable magnetic materials; the hunt
for proper materials and sizes had begun.

On August 1, Forrester talked over the phone with Dr. Eberhard
Both, a specialist in the subject working at Fort Monmouth, New
Jersey, about the composition of Deltamax and other materials, and
during August and September he was carrying on conversations
regarding his need with Allegheny Ludlum and its subsidiary,
Arnold Engineering. In the fall, Forrester selected William N.
Fapian, a graduate student looking for a thesis topic in the
Laboratory, to go to work testing individual cores by the dozen in
order to ascertain ranges of performance and to pick out cores
exhibiting exceptionally good properties.

To someone visiting the Laboratory it might have appeared that
young Papian was engaged in a routine chore of sorting and testing
the tiny rings, but both Papian and Forrester regarded it as a
development project committed to converting an attractive theo-
retical principle to reliable practice. "He set me to work

11.7

developing magnetic-core storage, 41 recalled Papian, "and he
treated me in that way anybody in the lab would recognize, support-
ing and guiding the effort to success with very little detailed

8
nagging." During October new cores received from Allegheny

Ludlum proved to be some three hundred times faster in their switch-
ing interval from one state to the other: down from ten thousand
microseconds to about 30 microseconds . Electrostatic tubes were
still faster, however.

Following his practice of keeping in touch with the work the
Laboratory engineers were carrying on and carrying out his speci-
fied responsibility for the graduate assistants as they pursued
their thesis investigations, Everett made it a policy to get
around the Laboratory every couple of weeks, stopping to go over
and discuss the work in progress. Thus, he was aware both of the
details of Papian' s work and of the status of the electrostatic
storage tube program. In response to a request from Forrester, he

prepared a cost estimate early in January 1950 on the storage

9
tubes, and Hugh Boyd followed this with another in mid-February.

In the summer of 1949 electrostatic storage tube research and

development had not yet reached the stage at which assemblies were

being built to be incorporated into Whirlwind I. There was no

question that an array of tubes could be built and installed into

the computer as its automatic, internal "memory," but reliability

problems and opportunities for improvement continued to present

themselves.

11.8

Meanwhile, the other elements of the giant computer, rack on
rack and row after row, were being steadily built up, tested, inter-
connected, tested, further interconnected and tested again, and it
would not be long until Whirlwind I would be in existence as a com-
puter without any storage facilities more ample than the small,
relatively static, hand-test storage designed to check the opera-
tion of portions of the machine and the steps in preliminary
operations, the success of which would render the machine ready
for larger and faster internal storage.

Since Forrester had no intention of allowing Whirlwind to
become a computer with a "white -elephant" reputation, lacking fast
and ample storage, he took care to assess and reassess his situa-
tion. But his own and Papian's investigations showed that the
attractive promise of the cores could not be realized in time to
meet the needs of an otherwise operational Whirlwind. Even before
this information was well in hand he had told ONR in his fall
Summary Report that "the next quarter will be used to construct the
first set of tubes for WWI." His sense of engineering prudence
caused him to qualify this bold assertion, however: "the electro-
static storage system will not be connected to the rest of the
computer until it has been demonstrated that trouble-free operation
can be expected.... Reliability runs of the complete storage row,"
he concluded, "are not expected until February 1950."

Boyd's February cost analysis of proposed "standard, 100-series"
electrostatic tubes, each storing 256 bits of information in a

11.9

two-dimensional 16 x 16 array, indicated a high cost of approxi-
mately $1500 per tube and a low rate of production of "one and

12
one-half satisfactory tubes per week." The cost was indeed high,

but it was the best that could be done, and under the circumstances

must be accepted, even though the cost (including all tubes) per

tube per week to the Laboratory approached $2250.

Necessity alone is never the mother of invention, yet the
relatively marginal viability of the electrostatic tube as a
research and development product was clear to Forrester. He sought
to put the best face on the situation he could in reports he sent
to MIT Provost Julius A. Stratton's office in March of 1950, in
September, and in the following January, 1951. In none of these
formal letters did he inject a word about the magnetic-core research
then being carried forward. He was not making a secret of this
research; on the contrary, he was keeping Valley and others on the
MIT staff informed in an informal manner as 1950 wore on. He
simply was unwilling to speak of his core-storage concept in the
same breath with descriptions of the nearly-operable, newly-
operable whirlwind I computer.

Three days before the fateful meeting with ONR representatives
at which Valley stepped in formally to offer new federal funds,

Forrester wrote Stratton's office that "we expect to have the first

13
bank of storage tubes operating before the first of July,..."

He offered this estimate in the context of remarks opining that "we

are probably being much too modest in our claims for the machine."

11.10

Certain minimum terminal equipment, together with the first bank
of tubes, would give by the end of summer "a computer even more
complete and flexible (with respect to the entire range of possible
computer applications) than any other computer either now in exis-
tence or with prospect of materializing within the next two or
three years . "

He was indeed putting the best face on the situation, perhaps
too enthusiastically, for he went on to say, "It will be better for
some applications than others. It will be the only computer which
can be applied to many important problems (because of its speed) .
In a few categories of applications it may not be (in its initial
form) as useful as some other machines. The machine at that time
should no longer be the limitation on advancement Of digital com-
puter utilization; the shortage of enough trained personnel will
become the bottleneck."

These were brave (and prophetic) words, drafted in full aware-
ness of the impending meeting with ONR over funding problems — the
meeting at which Valley stepped forward- — , but they were no solution
to the internal-storage problem. Forrester himself admitted that
one bank of 256-digit storage tubes would provide only one-eighth
of the ultimate storage capacity he then had in mind. Unwilling to
leave his terms "minimum" and "ultimate" undefined, he pointed up
his meaning in typical personal style. Were a stranger to ask,
"Can you do my computing job?" wrote Forrester, the answer appropri-
ate to a minimum computer's capacities must be, "Probably, but we

11.11

must analyze it to find out." But if one possessed the ultimate
system Forrester had in mind, then the answer to that question
"can fairly safely be, 'Yes, what is it?"* 15

But the storage tubes were not part of the computer yet, nor
were they operating in July, as he had hoped, nor in September.
"The final testing and alignment of the first storage bank is mov-
ing along steadily," he wrote, injecting first a diplomatic note
of tempered optimism, "but more slowly than I had expected," he

added, perhaps ruefully, perhaps philosophically, but nonetheless

_ - ^, 16 He was finding it a touch-and-go-business, and
matter-of-factly. . e

a careful reading between the lines, augmented by the power of hind-
sight, indicates again how heavy were the pressures that weekly
were becoming heavier as the Air Force phase of proposed application
began to take more explicit form. Careful testing of storage tubes
and associated circuits in each digit column was producing "numerous
small difficulties;" this was to be expected with new equipment,
Forrester argued. But "our biggest problem," he confessed, "is
that of trying to do in two months what we have had six months to
do on other comparable parts of the computer."

They were running out of time, including the planning time
Forrester had allotted himself two years earlier. In the summer of
1948 he had published a "Long Term Whirlwind Schedule." It has
appeared in the August Summary Report to the Navy. As of that date,
Forrester had spent thirteen months, by his own reckoning, on the
analog-computer phase of the Aircraft Stability and Control Analyzer

11.12

before abandoning that line of inquiry at the end of 1945. From
about October 1945 to February 1946 he was in transition, and from
January 1946 he and the engineers he gathered around him were com-
mitted to the digital computer as the most practical alternative.

Digital design studies involved the Project for the following
ten months of 1946, and during the eighth and ninth months specifi-
cations of the speed and storage requirements of the ASCA had been
settled upon. Speed and efficiency considerations during the ninth,
tenth, and eleventh months led to the decision to undertake parallel,
or simultaneous, digit transmission, and by the end of 1946 it was
clear to the Project Whirlwind engineers that they were talking
about a machine that could cope with far more than aircraft analyzer
problems; they were undertaking to develop a practical, general-
purpose machine, or at least the prototype of one.

During the first half of 1947, the block diagrams detailing
the basic logical (though not electronic) functions for a parallel
arithmetic element and central control had been worked through by
Everett, and in the spring of that year the design of the Whirlwind
prototype had begun to be laid down. Since the proper operation of
the arithmetic element was utterly essential to the success of the
machine, the five-digit multiplier was conceived as a preliminary
test component and committed to construction during the middle of
1947.

Sixteen months after they began the design of Whirlwind itself,
they began to receive electronic panels from Sylvania under subcon-
tract, and by the end of 1948 the basic racks were up, the arithmetic

11.13

element of the computer was going in, and power sources were being
installed. So in 1948 the physical computer began to emerge,
acquiring the physical equipment of the central control element
during the first half of 1949 and beginning to receive minimum
input and output equipment after mid-1949 <,

During the third quarter of 1949, "after nearly two years of
research, design, construction, and tests," Forrester was happy to
report, "the computing section of WWI, has just passed a most sig-
nificant milestone: solving an equation and displaying its solu-

2 3
tion" on an oscilloscope, showing values for x, x , and x .

Previous test problems had called for only "single-point solutions,"

whereas the progressive display required by this problem, "no

matter how simple, can result only when all the basic parts of the

18
computer act in harmony."

Only the storage element was left, and the state of completion
of the rest of the computer, as 1949 passed into 1950, made it
mandatory that it begin to be incorporated in order that testing,
maintenance, checking, and trouble- location procedures might con-
tinue. These had begun to be applied to the emerging machine at
the start of 1949 and were essential prerequisites to a fully
tested and operable machine — operable, that is, for the purposes, of
putting it to work and thereby exploring its nascent capabilities.

But the incorporation of the storage element depended upon the
state of progress of the storage- tube research and development which
had been carried forward since 1946, especially after parallel
transmission of digits had been decided upon late in 1946.

11.14

By September 1950 Forrester was still able to point out, as
he did in his letter to Stratton, copies of which went to Valley
and to C. V. L. Smith of ONR, among others, "Thus far at least,
the computer status is not holding up progress of the Air Force
intercept program, although a working storage-tube bank will soon
be necessary.

In the meantime, the test storage had proved capable of meet-
ing the demands placed on it in "testing out the terminal equipment
and telephone line transmission of radar data from Bedford.... We

have made two trials of transmitting radar data into and through

20
the computer with promising results."

While work on the first bank of storage tubes continued in

the Barta building, Papian reported in September on his work of the

preceding nine months. "The problem is bracketed on the one hand,"

he wrote, "by a metallic core . . . which has excellent signal

ratios and a 20 micro-seconds response time, and on the other hand

by a ferritic core . . . which has only fair signal ratios and a

21
1/2 micro-second response time."

Prospects were sufficiently attractive so that Papian proceeded
to a 2 x 2 planar array of cores. In October he obtained "success-
ful operation . . . with 30 micro-second switching time," demon-
strating experimentally one of the conceptions Forrester had set

forth in a laboratory report he duplicated in May and submitted in

22
June to the Journal of Applied Physics as an article.

11.15

While this work was progressing, MIT prepared a proposal in
October 1950 to present to ONR regarding research problems con-
sidered worth pursuing not by Project Whirlwind alone but by
another laboratory. The title of the proposal was "Program of
Applied Research at the Laboratory for Insulation Research," and
the first project proposed was one urging ONR to support the work
of Dr. von Hippel in conducting "an investigation of the ways and

means by which the hysteresis loops of ferroelectrics and of ferro-

23
magnetic semiconductors can be shaped to order." The pertinence

of the work to progress in computer research and development was
stressed: "The present-day ferrite materials show that this goal
can be achieved by dielectrics, if rectangular hysteresis loops
can be produced. Project Whirlwind is therefore extremely interes-
ted in the outcome...."

They had reason to be. For while the magnetic cores were
becoming increasingly attractive prospects for further research,
the electrostatic tubes were posing new problems. By the middle of
1950 the smaller tubes, equipped with guns providing a shorter
"throw," clearly were not living up to expectations or specifica-
tions: "... a reliable operating tube with 32 x 32 density has

not been achieved." Thirteen of thirty- three tubes produced in

25
one three-month period were research tubes. When the first bank

of standard tubes was connected into the computer in July and short

programs were run, successful operations could be obtained for

several minutes at a time, or even for an hour, but the tubes were

11.16

not reliable. "... the behavior of the storage," said the third
quarterly report for 1950, "depended on the programs used and
their frequencies, and it varied when different areas of the storage
surfaces were used. There was evidently much that we did not under-
stand about the operation of the storage as an integrated part of

•i26
the computer.

It was in this same fall report that the possibilities of

magnetic-core storage were first discussed in some detail in a

periodic, formal, progress report for the record to the Navy. The

results of Papian's work were presented and the proposed direction

of further investigations was indicated. Two lines of inquiry

appeared desirable, one devoted to "improving materials to reduce

eddy-currents and increase hysteresis- loop rectangularity," and

the other aimed at "uncovering and solving the problems associated

with operating large numbers of these cores in a high-speed memory

- 27
system.

The core-storage article Forrester had sent to the Journal of

Applied Physics in June appeared in the issue for January 1951.

Also during that month Forrester made another formal report to

28
Stratton's office. In it he reported that since the first bank

of storage tubes had been connected (in July) the computer had been

"operating satisfactorily much of the time." In consequence, they

had been able to make "steady progress" on their Air Force task and

29
at the same time get "organized for engineering applications...."

11.17

The storage tube problem was not one to be solved In a hurry,
because the complex structure and operations of the tubes posed so
many sub-problems. Forrester felt compelled to admit at the
beginning of 1951 that "performance of the storage tube bank is
not yet good enough to permit predicting future scheduled perfor-
mance of the machine with complete confidence." Again he made no
reference to the possibilities of magnetic-core storage.

He turned instead to the matter of scientific applications of
the machine, a goal which Mina Rees had undeviatingly held in view
while ONR was bringing Whirlwind expenses into line with those of
similar funding operations which ONR sponsored. Comments that
Dr. Rees had made to Professor Morse caused Forrester to describe
the resources of the four-man mathematics and applications group
on the Project staff, headed by Charles Adams. While most of the
work of Adams ' group had been devoted to working out "basic machine
techniques and procedures, .. .a few specific problems" were appro-
priate to call to the attention of Stratton's office in order to
show that the computer was indeed moving steadily toward being a
useful tool.

The proper integration of output equipment and solving the
electrostatic storage problems, including increasing the storage
capacity beyond the present, unreliable, single bank, remained as
tasks that would not be concluded by the end of the 1951 fiscal
year, as Forrester had hoped. Money problems were by no means over,
just because ADSEC had come to the rescue a year ago.

11.18

In the preliminary funding discussions during that January
twelve months earlier, when Valley and his colleagues had for the
first time visited Project Whirlwind, John Marchetti of the Air
Force Cambridge Research Laboratories had tentatively suggested an

annual expenditure rate of $200,000 from Air Force funds to under-

30
write Whirlwind's participation in the committee's program. By

March 1950, this amount had more than doubled, implying the urgency

of the program and the importance the committee had come to attach

to the role of Whirlwind I. In addition to the figure of $500,000

which Valley mentioned at the March discussions with ONR, the Air

Force the following month redirected the balance in the air traffic

control study contract to "the experimental application of digital

31
computers to air defense in accordance with the needs of ADSEC."

This action apparently caused some confusion concerning the

amount of money the Air Force would make available. The MIT Provost

understood the amount would be some $600,000, including $120,000

left in the Air Traffic Control study account, whereas John Marchetti

believed the $600,000 to be an additional sum. Marchetti 1 s projected

budgets for the following two fiscal years also included an annual

allocation of $600,000, again effectively demonstrating the com-

32
mittee's reliance upon Whirlwind I. The actual amount transferred

by the Air Force in November was $480,000; this, plus the $20,000

given by ONR and the $120,000 from the Air Traffic Control study

account, made initially available for the ADSEC phase of Project

Whirlwind a sum of about $620,000.

11,19

The urgency which the Air Force attached to the work of the
Valley Committee was illustrated not only by the redirection of the
Air Traffic Control study, but also by the manpower and financial
requirements imposed upon the Cambridge Research Laboratories to
support the committee's efforts. Despite the reluctance of his
laboratory chiefs, John Marchetti, without doubt under pressure
from higher echelons, established within the Laboratories an
"interim Air Defense Group," possessing a "priority in excess of

any other job in the house." In his directive, Marchetti noted

33
that "the task must be done, and no further delay can be tolerated."

In addition to drawing upon the best scientific and engineering

talent in the laboratories, the Air Force, in finding funds for the

Valley Committee, also tapped the Laboratories ' appropriation, again

34
to the dismay, if not the irritation, of many CRL personnel.

In April of 1950 in accordance with its instructions from the
Air Force, Project Whirlwind discontinued its work on the air traf-
fic control problem and turned its attention to "the application of

high speed computers to tracking while scanning in accordance with

35
the needs of the Valley Committee." This "TWS" phase of Project

Whirlwind's activities was continued under the air traffic control

study contract until the latter's termination on June 30, 1951, at

which time Project Whirlwind's participation was financed through

a new contract, AF 19(122)-458, administered by the Institute as

account number DIC-6889. The new contract reflected, of course,

official confirmation of the agreements reached by MIT and the Air

11.20

Force the previous January when the Institute agreed to implement
the ADSEC recommendations, conduct further investigations into the
air defense problem under the code name "Project Charles," and
establish Project Lincoln. All three programs were initially
coordinated and supervised by the first director of Project Lincoln,

Dr. F. Wheeler Loomis, on leave to MIT from the University of

.... . 36
Illinois .

Since Whirlwind I was still under construction during 1950,

the initial work carried out for ADSEC "consisted of (1) studying

the TWS problem in order to program (or 'instruct') the computer and

(3) devising a means of inserting radar data into the machine." The

radar set which the group anticipated using was a NEW set located

at Hanscom Air Base in Bedford. This set had previously been used

by the Cambridge Research Center to test one of its developments,

a Digital Radar Relay, a system which "permitted transmission of

37
range and azimuth data over an ordinary telephone line." Once

the components were ready and the system was joined, experiments

were conducted which produced flight data. But the data were of

poor quality because of erratic behavior on the part of the MEW

radar. So between February 12 and March 12, 1951, the system was

38
shut down for overhaul and repair.

Once back in operation, the system received further minor
technical corrections until it successfully "tracked a single air-
craft and computed the proper magnetic heading instructions to
guide the aircraft to an arbitrarily chosen geographical point."

11.21

Subsequent similar successes led to an attempt to perform "a
computer-controlled collision-course interception." On April 20,
1951, three such interceptions were successfully completed. The
pilot of the intercepting aircraft reported that from a distance

of about forty miles, he was brought to within 1,000 yards of his

39
target on each occasion.

The interceptions were hailed as a success. They demonstrated
the feasibility of the concept, they vindicated, in a preliminary
way, the predictive arguments Crawford, Forrester, and Everett had
held to for years, that such practical, real-time operations lay
within the power of electronic information systems, and they pro-
vided the basis for an expanded, experimental, multiple-radar sys-
tem to be established later in the Cape Cod area. At the same
time, such tests offered "valuable experimental experience in
preparation for operating" Whirlwind I in the expanded system.

The successes achieved by ADSEC and Project Whirlwind won the
commendation of Air Force Chief of Staff Vandenberg. In a letter
to George Valley, he observed that the "successfully accomplished
digital computation of interception courses with the Whirlwind
Computer" gave "real hope of being able to eliminate some of the
delays and inaccuracies inherent in conventional manual Air Defense
control systems."

There are as many ways of determining when a computer is
"in operating condition" as human ingenuity can devise, because the
digital computer in an integrated system of electronic circuits

11.22

which evolves through many stages of testing and preliminary elec-
tronic operation of, first, its elements, then groups of its
elements, then eventually all of the elements together. Further,
there is the repetition of this pattern, once the entire machine
is considered operable, through many levels of logical complexity
as determined by the programs and information-processing paces
through which the machine is put. However, long before the elec-
tronic capacities of the components of the machine have finished
being checked out, the logical calculational capacities have begun
to be explored, inasmuch as these logical capacities take expres-
sion first in the electronic operations of the machine.

If one is most concerned about whether a computer will per-
form the basic arithmetical operations, then these may become his
principal criteria of its operability. If a given storage capacity
must be demonstrated first, then that is the consideration which
may determine whether a particular computer is "really" in opera-
ting condition or not. If a computer is expected to carry out a
certain type of program, then that may become the standard of its
"true operating condition." Nor are these the only examples.

Forrester, Everett, and their engineering staff recognized
full well that the usefulness of this computer which they had been
touting as a "general-purpose" machine depended on the size of its
internal, automatic storage. The manually controlled test storage
could be set to put the other control and calculational elements of
the machine through enough paces to demonstrate their capacities,

11.23

their speed, their reliability, and in this respect, determine
whether they would "really" operate or not. But an operating
standard which could only be demonstrated with the passage of time
was that of how dependably the machine would perform its multiple
operations and how often it would have to shut down, or even pause,
for repairs .

The Project engineers and many of the technicians had suffi-
cient evidence, and from that evidence, knowledge that the computer
would calculate as they had expected it to long before the intercept
tests were run. They did not know how it would operate with full
storage capacity because such capacity had not yet been achieved.
Meanwhile, the success of the first air-intercept tests was another
milestone, and a dramatic one, demonstrating the potentiality of
Whirlwind as a truly useful machine.

In this perspective, the chronic troubles the Project had been
having with the electrostatic storage tubes were not crippling, for
those troubles had not prevented the tubes from contributing to the
success of the intercept tests of the spring of 1951. Since June

1950 the first bank of 16 tubes, comprising a storage capacity of

42
256 registers, had been connected to the rest of the computer.

Preliminary results were at once gratifying and irritating. The

tubes worked, but not all of their storage capacity was dependable.

Parts of the storage surfaces would operate without error for over

an hour, while other parts would not hold up beyond five minutes.

It was the over-all delicacy and instability of the tube-bank's

11.24

performance that caused Forrester, Everett, and their staff to
remain profoundly dissatisfied with the reliability prospects.
These were quite intolerable when measured by the standards origin
nally laid down. Improved quality control in tube design and tube
manufacture was imperative.

However promising magnetic-core storage appeared to be, the
research and development problems involved in magnetic core appli-
cations were altogether too formidable and time-consuming to meet
the current needs of Whirlwind, for it was now a computer going
through shake-down operations . So there was nothing for it but
to continue testing and adjusting present tube and circuit designs
while continuing also to construct and test special research modi-
fications on tubes not wired into the rest of the computer. Obvi-
ously, the key lay in controlling the characteristics of the indi-
vidual tubes that were to serve as the building blocks of the total
internal storage element.

During 1950 and 1951, then, the Project was engaged in explor-
ing and improving the operating capacities of the entire computer
with one hand, so to speak, while carrying on development work on
component elements with the other hand. The principal elements
needing improvement were the internal storage and the input-output
equipment. Aware that the possible modes of input and output were
many and varied, and estimating at the start that the design and
development problems were far less formidable with regard to this
terminal equipment than with regard to over-all machine control,

11.25

arithmetic computation, and internal storage, Forrester had been
willing to let the Special Devices Center contract for part of the
terminal equipment separately. After he had participated in estab-
lishing the requirements this equipment must meet in order to
integrate compatibly with the rest of the computer, he had been
satisfied to have the Eastman Kodak Company take on the job of
designing and developing the reading and recording unit. Eastman
would work with Project Whirlwind at the engineering level but
would be responsible contractually, fiscally, and administratively
to the Special Devices Center. The input-output control element
would be developed at MIT, in order to ensure that numbers would
be transferred satisfactorily between the input-output register of
Whirlwind and whatever external memory and recording devices were
used- -in this instance, the Eastman reader-recorder.

The Eastman Company had delivered the first reader-recorder

unit on September 13, 1949, after testing it during the preceding

43
three months. Reliability tests were next in order, to find out

how well and consistently the unit would "record binary numbers on

photographic film by... light from... a cathode ray tube," and how

well and reliably it would "read such recorded data by scanning of

44
the processed film with a cathode ray tube."

By the beginning of summer, 1950, the reader-recorder unit was

proving sufficiently unreliable when connected to the computer to

indicate the need for more extensive testing than had been planned.

Characteristically, the Summary Report issued that spring began

11.26

discussing other forms of input-output equipment: "The input-
output requirements have been studied," said the report, "and a

proposal has been made for the minimum amount of terminal equip-

45
ment needed to make the computer useful in handling general problems."

Input equipment called for included a typewriter, a coded-paper-tape
punching device, and film units. Output equipment would include a
typewriter, a paper-tape punch, film units, and "a versatile oscil-
loscope display." Acquisition and integration of such units would

-u • 46
occupy the coming year.

By the time of the successful air intercepts over the Atlantic
Ocean in the spring of 1951, the Eastman units had been quietly dis-
carded, a Flexowriter typewriter and punched- tape units were in
service, and plans were laid for "a flexible system that will
accommodate various kinds of terminal equipment," including paper

tape, magnetic tape, oscilloscopes, a scope camera, and a control

47
to introduce preprogrammed marginal checking whenever desired.

More important than the incorporation of improved input-output
equipment was the progress achieved by the beginning of 1951 in
electrostatic storage tube design. As a result of this advance,
Forrester felt free to proclaim to ONR in his official report of
the spring of 1951 that Whirlwind had become "a reliable operating
system." By the end of March, 1951, applications were being pro-
grammed on the machine a scheduled thirty-five hours a week. "Of
this time, 90 percent is useful." The computer had achieved as

many as seven consecutive hours of error-free operation more than
48

11.27

Alterations in the "charge on the glass windows" of the elec-
trostatic tubes were identified as. a major source of trouble and
led to a redesign, the 300-series tube, in which coating "the entire
inside of the envelope. . .with aquadag" removed the fluctuations of
the high-velocity gun's beam that had earlier caught the investi-
gator's attention. Once all sixteen tubes had been replaced,

Forrester was willing to call the computer "a reliable working

49
system. Five years had passed since Forrester had made the

decision to abandon the ana log- computer approach to the Aircraft
Stability and Control Analyzer problem and had committed himself to
the alluring but virtually untried electronic digital computer.

A final significant footnote: the regular vacuum tubes
employed in the pulsed circuits had become so reliable after the
interface problems had been solved that wholesale replacement of
tubes within less than 20,000 hours was considered to be not only
unnecessary but "unwise." A new high in reliability and longevity
had been achieved in the course of Whirlwind's development, poten-
tially increasing the life span of many standard-design vacuum
tubes from a few hundreds of hours to several thousands of hours.

NOTES TO CHAPTER 11

1. Testimony of J. W. Forrester in records of Patent Interference
No. 88,269, pp. 27-28.

2. Ibid , p. 31.

3. J. W. Forrester, Coincident-Current Magnetic Computer Memory
Developments at MIT , p. 2. This paper was given at the Argonne
National Laboratory Computer Symposium of August 4, 1953. Cf .
J. W. Forrester Memo. No. M-70, sub j : "Data Storage in Three
Dimensions," April 29, 1947.

4. Project Whirlwind Summary Report No. 17 . February, 1949, p. 7.

5. Ibid , p. 8.

6. Interview, J. W. Forrester by the authors, July 24, 1964.

7. Interview, R. R. Everett by the authors, July 31, 1963.

8. Letter, W. N. Papian to T. M. Smith, February 12, 1968.

9. Memorandum, R. R. Everett to J. W. Forrester, January 9, 1950;
Memorandum L-18, H. R. Boyd to J. W. Forrester, February 15,
1950.

10. Summary Report No. 20, Third Quarter. 1949 , p. 17.

11. Ibid , p. 27.

12. Memorandum L-18, February 15, 1950, p. 1.

13. Letter, J. W. Forrester to J. A. Stratton, March 3, 1950, p. 1.

14. Ibid .

15. Ibid , pp. 3-4.

16. Letter, J. W. Forrester to J. A. Stratton, September 28, 1950,
p. 1.

17. Ibid .

18. Summary Report No. 20, Third Quarter, 1949 , p. 9.

NOTES TO CHAPTER 11 (CONTINUED)

19. Letter, J. W. Forrester to J. A. Stratton, September 28, 1950,
p. 2. The upper right-hand corner of page 2 bears the date
"Sept. 23, 1950."

20. Ibid .

21. Report R-192 by W. N. Papian, sub j . : "A Coincident-Current
Magnetic Memory Unit," September 8, 1950, abstract.

22. Report R-187 by J. W. Forrester, subj . : "Digital Information
Storage in Three Dimensions Using Magnetic Cores," May 16, 1950.

23. W. N. Papian patent interference testimony. The proposal
pointed out that this problem was "at present in the center
of interest for highspeed digital computers like Whirlwind.
Here simple yes-no devices are required that can switch in
fractions of a microsecond and lend themselves to a three-
dimensional storage of information."

24. Ibid .

25 o Summary Report No. 23, Second Quarter, 1950 , p. 15.

26. Summary Report No. 24, Third Quarter, 1950 , p. 6.

27. Ibid , p. 24.

28. Letter, J. W. Forrester to Prof. J. A. Stratton, January 18,
1951.

29. Ibid , p. 1.

30. J. W. Forrester, Computation Book No. 49 , p. 84, entry for
February 1, 1950.

31. C. R. Wieser, Lincoln Laboratory Quarterly Progress Report ,
Division 6--Digital Computer , June 1, 1952, pp. 6-9.

32. J. W. Forrester, Computation Book No. 49 , p. 102, entry for
April 12, 1950.

33. Directive, John H. Marchetti, Director, Radio Physics Research,
to Col. Mitchell, R. E. Rader, Dr. Hoi lings worth, Dr. Spencer,
Dr. Samson, Dr. Foster, and Mr. Davis, June 29, 1950.

NOTES TO CHAPTER 11 (CONTINUED)

34. Memorandum, Robert E. Rader, Chief, Air Defense Group, Base
Directorate, Radio Physics Research, sub j . : "Transfer of
Personnel to Air Defense Group," July 5, 1950; Memorandum,
R. E. Rader to Dr. George E. Valley, October 31, 1950.

35. Servomechanisms Laboratory, MIT, Summary Report No. 6 ,
April 25, 1950--July 25, 1950 , "Submitted to Watson Labora-
tories, Air Material Command, Under Contract AF 28 (099) -45,"
p. 2.

36. Conclusions--Scientific Advisory Committee Meeting, January 19 ,
1951 ; Contract AF 19 (122) -458, January 20, 1951. Memorandum
A-117, H. Fahnestock, Sub j . : "New Funds under Contract AF 19
(122) -458, DIC 6889," May 7, 1951.

37. C. R. Wieser, Lincoln Laboratory Quarterly Progress Report ,
Division 6--Digital Computer , June 1, 1952, pp. 6-9.

38. J. W. Forrester, Computation Book No. 49 , pp. 135 and 136,
entries for February 1, 1951; Servomechanisms Laboratory, MIT ,
Summary Report No. 9, January 25, 1951- April 25, 1951 ,
"Submitted to Air Force Cambridge Research Laboratory under
Contract AF 28 (099)-45," p. 1.

39. Ibid , p. 23; C. R. Wieser, Lincoln Laboratory Quarterly
Progress Report, Division 6 — Digital Computer , June 1, 1952,
pp. 6-10; Memorandum M-1515, D. R, Israel, sub j . : "Intercep-
tion Experiments with Bedford Mews," June 11, 1952.

40. Memorandum M-2092, C. R. Wieser to J. W. Forrester, sub j . :
Experimental Interceptions with Bedford Mew Radar," April 23,
1951.

41. Letter, Hoyt S. Vandenberg, Chief of Staff, USAF, to George E.
Valley, Jr., Chairman, ADS EC, May 28, 1951.

42. Summary Report No. 23, Second Quarter, 1950 , p. 5.
43 o Summary Report No. 20, Third Quarter, 1949 , p. 28.

44. Ibid .

45. Summary Report No. 22, First Quarter, 1950 , p. 21.

46. Ibid.

NOTES TO CHAPTER 11 (CONTINUED)

47. Stannary Report No. 25, Fourth Quarter, 1950 and First Quarter ,
1950 and First Quarter. 1951 . p. 5.

48. Ibid , p. 6.

49. Ibid , p. 5.

50. Ibid.

Chapter Twelve

MAGNETIC CORES AND R&D PROGRESS

One of the follow-on test projects that the investigators
working in Project Charles had recommended was the more elaborate,
Cape Cod multiple-radar system- Whirlwind I would form the machine
element at the information-processing, command-and-control center
of this system. The final report of Project Charles, although more
comprehensive and detailed, recommended a solution of the air de-
fense problem which was essentially similar to that proposed by the
Valley Committee, the feasibility of which was being investigated
and demonstrated by Whirlwind I. Project Charles' investigators
were familiar, of course, with the successful experiments conducted
by Project Whirlwind and ADSEC. They had been unable to find any
"other solution to air defense data processing . . . equal in long
term value to the digital transmission and automatic analysis of
data." Hence they recommended the construction of a model system
in the Cape Cod region of Massachusetts. This system could consist
of a series of radar stations tied to Whirlwind I at MIT. The data
obtained from this experimental network, it was anticipated, would
provide guidelines for the design, development, and construction of
a more sophisticated digital computer. In this next-generation
computer there would be incorporated the unique functional charac-
teristics required by the information and control center of the
proposed centralized air defense system.

12.1

12.2

The model system built in the Cape Cod region became the joint
responsibility of Divisions II and VI of the Lincoln Laboratory,
the air-defense laboratory built and managed by MIT for the purpose
of coordinating and implementing the recommendations of ADSEC and

Project Charles and for performing within its own facilities

2
research, development, and tests in the general area of air defense.

The primary task which thus confronted Lincoln Laboratory at this

time was the "development of a system using a high-speed digital

computer to receive, process, and transmit air-surveillance, iden-

3
tification, and weapon- guidance information."

Concurrently with the establishment and organization of Project
Lincoln in 1951, MIT reorganized its computer program. At Jay
Forrester's request, Project Whirlwind was detached from the Servo-
mechanisms Laboratory and reconstituted as the Digital Computer

4
Laboratory under Forrester's direction in the autumn of 1951.

This administrative reorganization within the Institute at the same

time served to placate ONR and facilitate administrative relations

between Project Whirlwind and Project Lincoln.

In the fall of 1951 Jay Forrester discussed with Professor

Wheeler Loomis, Lincoln's director, the nature of the relationship

between the two groups. On this occasion Loomis mentioned two

possible administrative relationships: "one a sort of sub-contract

arrangement . . „ the other a closer administrative tie. ..."

Forrester, as could have been anticipated, noted his preference for

an independent status and suggested that the Digital Computer

12.3

Laboratory's part of Lincoln's program remain at Cambridge until
the computer his group was to design especially for air defense
had been assembled or put "into initial operation." This, he
estimated, would not be until 1954.

When Project Lincoln was first organized by MIT, responsi-
bility for digital computer research and development had not been
assigned to it. Second thoughts, however, added it to Lincoln's
program, and six months later those operations and staff members
of the Digital Computer Laboratory concerned with Whirlwind I and
its application to air defense were incorporated as "Division VI"
into Lincoln Laboratory. Consequently, Jay Forrester came to
wear two hats: one as director of MIT's Digital Computer Labora-
tory, the other as head of Lincoln Laboratory's Division VI. In
the latter capacity, he was responsible for Whirlwind's participa-
tion in the Cape Cod esperimental system and for the design and
construction of a digital computer possessing the characteristics
"desired for a future operational air defense system."

Forrester and his colleagues within Division VI had four major
tasks facing them from the outset: (1) the organization of the
Division and the formulation of guidelines for its relations with
its parent organization, Lincoln Laboratory; (2) the design, con-
struction, operation, and expansion of the Cape Cod system in coop-
eration with Division II; (3) the design and construction of a
prototype air defense computer and its ancillary equipment, with
the necessary concomitant research and development; and (4) the

12.4

selection of an industrial source for production of the air defense
computer. All four tasks were pursued concurrently during the
closing months of 1951 and throughout 1952-53.

Administratively, Division VI was divided into six groups,
each charged with a specific responsibility within the overall
program. Group 60, under the direction of Harris Fahnestock, was
responsible for administrative services. Group 61, under C. R.
Wieser, was responsible for the Cape Cod system. Group 62, under
Norman Taylor, bore the responsibility for the design and construc-
tion of the projected air defense computer. Group 63, under D. R.
Brown, was responsible for magnetic materials. Group 64, under
S. H. Dodd, was responsible for the maintenance and improvement of
Whirlwind I. And Group 63, under Pat Youtz, conducted the electro-

Q

static storage tube development program.

Later, the Sage System Production Coordination Office and the
FSQ-7 Systems Office were organized within the Division. The Pro-
duction Coordination Office maintained "liaison with industrial and
military organizations outside Lincoln" and acted "as the coordina-
tion agency for Division VI portions of SAGE system planning and
implementation, thus providing suitable direction and control of
the program with respect to Lincoln's over-all responsibility for
the system." The Systems Office maintained coordination with the
International Business Machines Corporation, which subsequently

became the manufacturer of the production model of the air-defense

9
computer, the FSQ-7.

12.5

To a very great extent Forrester and his engineers enjoyed
the same independence and freedom of action within Lincoln that
they had enjoyed when attached to the Servomechanisms Laboratory.
Although nominally attached to the Laboratory, Project Whirlwind
had been virtually self-sufficient and independent. This was
especially true after the Project had so expanded in size that it
became necessary for it to be physically dissociated from the
Servomechanisms Laboratory and to move to separate quarters in the
Barta Building. Here the Project established its own shops and
maintained its own service crews, an autonomy acceptable to Nat
Sage's Division of Industrial Cooperation, which administered the
contract. Thus, the freedom of action permitted by Nat Sage in
technical matters had also been permitted by Nat Sage in administra-
tive matters. The cumulative effect confirmed Forrester's desire
and belief that Project Whirlwind was actually, if not contractually,

e - 10

a free agent.

During 1951 and 1952, prior to the physical integration of
Project Whirlwind into Lincoln Laboratory, Forrester and Everett
worked with the parent organization as Forrester had previously
worked with ONR. No members of the Whirlwind staff, for example,
attended Lincoln's meetings. So from the start, Division VI estab-
lished a pattern of autonomous behavior within Lincoln Laboratory,
and it was a pattern which persisted throughout the Division's
association with Lincoln. Even after Forrester's departure in
1956, Everett as the new director continued the semi -autonomous and

12.6

highly individualistic role until the Division was separated from
the Laboratory in 1959 to form the MITRE Corporation.

When the Division finally moved physically to the Laboratory's
new quarters at Hanscom Air Base in the Bedford-Lexington area of
Massachusetts, the giant Whirlwind I computer was left behind in
the Barta building on the MIT campus. The Division took its service
organizations with it, however: the print shop, machine shop, pur-
chasing office, etc. These services duplicated those already pro-
vided by the parent Lincoln organization, but Forrester and his
associates refused to disassemble the smooth-functioning organiza-
tion they had created at MIT as Project Whirlwind. There were those
within Lincoln who were irritated by this display of independence
and saw it simply as a determination to continue going "first class"
regardless of the added cost. But there were also those who took
advantage of the efficiency which the additional facilities and
services offered. Among the latter was George Valley, head of
Division II, who when in need of immediate assistance would resort
to the Division's supporting facilities.

At the same time that Division VI remained as highly indivi-
dualistic as Project Whirlwind had become, continuing to do business
its own way, it worked effectively with the radar engineers in
Division II. The latter, under the direction of Valley, were
responsible for aircraft control and warning and possessed a
special competence and experience, which Whirlwind's personnel
lacked, to set up the preliminary, experimental, Cape Cod system
employing several radar stations to feed data into Whirlwind I.

12.7

This particular kind of practical application of the general-
purpose computer, of which Whirlwind I was the first example,
required dimensions and directions of technical electronic knowledge
and experience that Project Whirlwind personnel had never acquired
and very likely could not have mastered in the brief time that
development schedules allowed. The engineers took this situation
for granted, and there was from the start the view that such pool-
ing of resources and talents as Division II and Division VI pos-
sessed was the natural course to pursue. It was the efficient way
to proceed, and it became the way they all successfully proceeded.

In this respect then, Whirlwind was not ruggedly independent.
Rather, its independence manifested itself in the group's self-
resourcefulness and daily conduct of its affairs. Although later,
by the move from the MIT campus to Lexington, it lost the injection
of the rich new blood that had been contributed by the graduate
students, who had contributed so much to the elan and the resource-
ful operations of Project Whirlwind, the Project's way of doing
things had already become firmly established, and it persisted for
some time, even without a continuing supply of graduate students,

in maintaining its special character and vitality against the

11
pressures imposed by the more conventional organization of Lincoln.

Two of the four major tasks facing the administrators of

Division VI were of vital importance to the over-all program of

Lincoln Laboratory, for unless they were successfully completed,

the program could become a total failure. These two tasks were the

12.8

construction, operation, and evaluation in cooperation with Divi-
sion II of the Cape Cod experimental system, and the design and
construction of a next- generation high-speed digital computer that

would possess the characteristics "required for an operational air

12
defense system." The Cape Cod experimental and evaluation system

recommended by Project Charles was a natural extension of the
earlier system established for the tests run by ADSEC. The Cape
Cod system was designed to test a multiple-radar network linked to
Whirlwind I; the total system would collect and process data and
would guide and control defensive countermeasures taken.

As an experimental system, Cape Cod served several purposes:
it developed "system concepts for a high- track-capacity system;"
it permitted new components to be tested in an operating prototype
of the air defense system envisaged by ADSEC and Project Charles;
it furnished data and other information necessary to the prepara-
tion of "specifications for digital computers designed specifically
for air defense;" and it permitted verification of the "soundness

of the whole concept by experiments using live radar data and con-

13
trolling live aircraft."

Division VI 's responsibilities for the Cape Cod system included

"air defense center planning, automatic information processing

(including data screening and automatic tracking), the computation

of control orders for weapons, and the provision of the digital

equipment necessary in the air defense center." These responsi-

14
bilities fell mainly upon Groups 61, 64, and 65. When work upon

12.9

the system was commenced by the groups concerned, it was antici-
pated that the program would consist of three parts: "(1) Con-
struction and operation of a three-radar network; (2) construction
of a 14-radar network; and (3) planning for a future operational
system." The first part, which was scheduled for completion during
fiscal year 1952, called for expansion of the single-radar system
with which the Valley Committee had conducted its successful inter-
cept experiments in the spring of 1951.

Throughout fiscal year 1952 the primary objective was "aimed
at learning how to track an aircraft through a network of radars
having overlapping coverage," in order to gain the knowledge and
experience necessary to implement the second part, expansion into
a 14-radar network. The hope was that by July of 1952 the larger
network would be under construction and that in the course of the
year, Whirlwind I would be sufficiently improved technically, by
the installation of magnetic drums, to expand the computer's capa-
city to handle the requirements of the full Cape Cod radar net.
If this schedule were met, then it was anticipated that prior to
June of 1953 the Division would be able "to commence operation with
the 14-radar network with automatic data-handling capacity for data
screening, automatic track-while- scan, and the control of a large
number of aircraft." Realization of these plans was of extra-
ordinary importance to the total air defense project, for in addi-
tion to laying the groundwork for an expanded experimental system,
the initial efforts would permit the military to evaluate the con-
cept even as it was in process of implementation.

12.10

Military evaluation of the project was more important than it
appeared to be on the surface, for even while Project Lincoln had
been in the process of organization by MIT and the national mili-
tary establishment, another air defense program had been under way
at the Willow Run Research Center of the University of Michigan.
This was a program similar in goal, but different as to the pro-
posed method of attainment. The competition between the two pro-
grams could be "sticky," not solely because of the impact upon the
educational institutions involved, but also because the situation
reflected the competition between the Rome Air Development Center
at Rome, New York, and the Air Force Cambridge Research Center at
Cambridge, Massachusetts. Both were Air Force agencies, but both
were striving to become preeminent if not dominant in air-defense
research and development and in the concomitant area of military
electronics. Furthermore, the competition was not divorced from
politics. Political representatives of the two regions concerned
sought to keep the program, since it held great potential for
economic growth and stability. Industry also did not remain
uninvolved, although the concern there was less narrowly regional.
The International Business Machines Corporation became involved
eventually in Lincoln's program. General Electric had expressed
interest in the program conducted at Willow Run.

The problems raised by the competing programs were finally
resolved in 1953 by the decision to terminate the University of
Michigan's program and concentrate the whole effort upon the

12.11

centralized concept then under development by Lincoln Laboratory

18
and the Cambridge Research Center.- There were the unavoidable

mutterings that the decision had been politically motivated, and

it would be naive to assume that political and economic pressures

played no role. Nevertheless, the feasibility of the concept

Lincoln was implementing had been demonstrated by the experiments

first conducted for ADSEC and then expanded under Divisions II and

VI of Lincoln Laboratory in the Cape Cod system. In comparison,

19
the Michigan system had "not been demonstrated as successful."

By March of 1953, the Cape Cod system, though incomplete, was
supplying "valuable experimental data from existing equipment."
The nucleus of the system was the network which had been put to-
gether for the ADSEC experiments, but two smaller radars located
at Rockport and Scituate, Massachusetts, using s lowed-down- video
data transmission links, had also been used in some tracking tests.
The data generated by the radars were fed into Whirlwind I. The
computer processed the information and provided "vectoring instruc-
tions for mid-course guidance of manned interceptors and ...
special displays for people who monitor and direct the operation of
the system."

The information and experience gained from such tests, using

"live aircraft, live radar, and an operating computer" proved

20
immensely "valuable in planning an advanced air defense system."

By December of 1953, the system was operating with a large radar

set (FPS-3) located at Truro and two gap- filler radars. The data

12.12

provided were processed automatically by Whirlwind I, which was

able to "carry the tracks of as many as 48 aircraft," and guidance

21
instructions were transmitted to intercepting aircraft. Although

by December, the "electronic systems and the programs" were func-

22
tioning smoothly, good radar data were still lacking.

The role Whirlwind was able to play in these exercises by the
end of 1953 hinged crucially upon the progress that had been made
in magnetic-core development work since Papian had built his first
2x2 planar array of cores in the fall of 1950. During 1951,
efforts were continued to obtain core material that would hold its
polarization in spite of electronic system "noise" and that would
switch rapidly from one state to the other when required. A 16 x
16 array of cores in one plane was constructed in order to test and
observe the effects of "noise," of the switching of nearby cores,
and the running of program patterns through the array. By the end
of 1951 "a fair demonstration of the practicability" of the arrange-
ment had been achieved: "error-free operation for periods of con-
siderable differences in the characteristics of the 256 cores used

23
in the array. The cores continued to appear promising, indeed,

yet they were still far from achieving the operating standards

required.

The addition of a second bank of electrostatic storage tubes

to Whirlwind at this time increased its storage capacity by 1024

24
registers without adding to storage access time. But experience

during 1952 with the new tubes revealed that contradictory to

12.13

earlier judgments the internal-storage problem was not yet out of
the woods. The new bank of tubes possessed a larger storage
density of 32 x 32 registers, compared to 16 x 16 in the first
bank. Unfortunately, they were by no means trouble-free. New
"400-series" tubes replaced the 300-series tubes to no avail. By
April 1952 the decision was reluctantly made to hold up adding any
more banks of tubes of the new design, and immediate plans to
replace the 16 x 16 tubes in the original bank were suspended. New
500-series tubes were being manufactured as fast as possible, but
it would take time to increase production sufficiently to maintain
an adequate stock of reserves, even if these did prove reliable.
"A large fraction of Whirlwind operating time" was being devoted to
maintenance and special checking of the installed storage tubes.
Furthermore, the limited supply of replacement tubes available
indicated it would be risky to put Whirlwind on a 3-shifts-a-day

schedule, even though applications programs were stacking up as

25
the demands grew for more machine time.

In the meantime, Fapian and his assistants were constructing

and testing another 16 x 16 planar array of cores composed this

time of non-metallic ferritic material instead of rings of thin .

metal ribbon wound on itself to form a doughnut. In May of that

same year (1952) it became clear that the switching speeds were

approximately twenty times faster than with the metallic cores--

down to one microsecond or less. So promising was the performance

of the non-metallic ferrites by July that Forrester, Everett, and

12.14

their engineers made the decision to build 32 x 32 arrays and stack
them sixteen-high, in a true three-dimensional arrangement.

Since Whirlwind was by then in heavy demand for preliminary
Cape Cod and other applications, it would not be possible to use it
to test the 16,384-bit core memory, comprising 1,024 registers,
which was then being built. The design of Whirlwind's operating
units had long since become standardized, however, so the solution
most practical to the engineers was to turn out a semi-copy of
Whirlwind — another computer to test the magnetic storage. So the
"Memory Test Computer" came into being as a concept during the
summer of 1952, and construction of this machine--more modest in
its size and capacities than Whirlwind--proceeded into the fall.
By November Forrester and Everett committed themselves fully to
the use of non-metallic ferrites for the Memory Test Computer's
storage bank. The following May 1953, the Memory Test Computer was
operating sufficiently well to demonstrate the "highly reliable
operation of a 32 x 32 x 16 magnetic ferrite storage."

To Forrester that summer the performance of the new core
storage meant the end of his search for a practical internal storage
element. Electrostatic storage tube manufacture and development
were abruptly halted as soon as it was clear that Whirlwind would
receive a core-storage unit- in- being, and on August 8, 1953, the
first bank of core storage was wired into Whirlwind. A second bank
of cores went in on September 5. Access time had dropped from 25
microseconds for tube storage to 9 microseconds for the magnetic

12.15

26
cores. But best of all, the chronic, incurable maintenance

troubles and the high costs of tube manufacture were at an end.
A brief valedictory in behalf of electrostatic storage was
nevertheless in order; the Summary Report pointed out that Whirl-
wind "could not have reached its present state of development with-
out ES. No other form of high-speed storage was available when

27
Whirlwind I was put in operation." Four years had passed since

Forrester had set Fapian to work on the magnetic core development
project, sorting, testing and analyzing the electromagnetic pro-
perties of the tiny rings .

Also during those four years Whirlwind had found a mission
and was about to spawn a successor that would take advantage of
advances in the electronics state of the art since Whirlwind had
been conceived. Long before the end of 1953 Forrester's engineers
had begun to consider the parameters the new defense computer
should have. The Cape Cod tests added valuable data and experience.
All of these would be of assistance in designing and constructing
the computer to be used in the projected continent-wide air defense
system.

Existing computers, including Whirlwind I, were "suitable for
studying the digital control of air defense," but they did not
possess the unique characteristics necessary to an air-defense
system. Moreover, the design and construction of an air-defense

computer was urgent; national security required the "availability

28
of an improved Air Defense System" as promptly as possible.

12.16

The question, before the end of 1951, was not "whether to build a

machine or not, but rather to build the best machine possible,

29
considering speed, cost, capacity, and complexity." In conse-
quence, concurrently with the construction and operation of the
Cape Cod system, Division VI, primarily through Groups 62 and 63,
embarked immediately upon a research and development program

pointed to the construction of an air defense computer, utilizing

30
primarily its personnel in Group 62 and Group 63. This computer

was first thought of as "Whirlwind II," then it became the "XD-1"

and finally the "FSQ-7."

In its approach to the research and development program which

resulted eventually in the FSQ-7, Division VI operated upon three

premises. First, Project Charles' recommendation for an "improved

air defense system using a digital computer information center"

had to be implemented and realized "as soon as practical." Second,

no contemporary digital computer could be more than a "laboratory

model" for the proposed system. An advanced design was needed

which would "improve reliability, reduce maintenance, be tailored

to the air defense application, and incorporate the necessary

facilities for the required terminal equipment." Third, the Digital

Computer Laboratory would furnish the "key personnel and background

31
experience for the estimated design program."

In a series of meetings held in the fall of 1951 to plan and

schedule the research and development program for the air defense

computer, those members of Division VI participating decided that

12.17

the computer should be fast, flexible, and reliable. It should be
as fast as, if not faster than, Whirlwind I. It should possess a
register length of 24 bits. The use of marginal checking, magnetic
cores, and transistors was considered in order to achieve maximum
reliability, even though sufficient perfection in the manufacture,
reliability, and techniques of using transistors and magnetic cores
still lay ahead at the time of the discussions. Reliability was so
important, the conferees believed, that they even considered the
installation of additional machines in the command and control
center, to be available as an instant reserve. A general-purpose

computer was essential, since flexibility was another major require-

32
ment. The time estimated to complete the program was three years.

By January of 1952 the design staff for Whirlwind II was in

process of organization. It was anticipated that the first half of

Fiscal Year 1953 would be spent in determining and establishing the

air-defense computer's characteristics and selecting its components;

the second half of the year would be devoted to its design. Four

major areas of concentrated effort were scheduled: "(1) a study of

new components and circuits, (2) the determination of optimum

machine logic to utilize these new techniques, (3) the development

of new magnetic materials for reliable high-speed storage and for

switching purposes, (4) close liaison with the Cape Cod System to

formulate the computer characteristics peculiar to air defense data

33
processing."

12.18

A problem of primary importance which beset the leaders of the
Digital Computer Laboratory and of Division VI from the outset and
which possessed special significance for Groups 61, 62, and 63
responsible for the Cape Cod system and Whirlwind II, was the
shortage of personnel trained and experienced in computer technology,
The problem was acute enough in itself, but it was further compli-
cated by the rapid physical and organizational expansion of the
Digital Computer Laboratory as it sought to meet its responsibili-
ties not only to the Air Force, but to the Navy and MIT as well,
for as Whirlwind I had become operational, programs other than
those in air defense were also assigned to it.

To cope with these complications as they bore upon Division VI,
Everett and Fahnestock recommended that regularly scheduled formal
meetings of group leaders and laboratory chiefs be instituted.
Such meetings, they reasoned, would keep the leaders aware of the
activities going on within other groups, permit critical analysis
of the program, and assist in the assignment of personnel and job
priorities. The Friday afternoon teas which in the days of Project
Whirlwind had provided those attending a pleasant and informal
means for keeping abreast of the program were no longer sufficient.
The responsibilities of the Digital Computer Laboratory were just
too complex, too great.

The first meeting held on March 26, 1952, considered the
seriousness of the shortage of experienced personnel and the impact
of the shortage upon the Whirlwind II and Cape Cod programs in

12.19

particular. Norman Taylor, responsible for Whirlwind II, predicted
that the schedule established for the design and construction of
the production model of the air defense computer would prove
"unrealistic," for of the twenty-three members of his group, only
four, excluding D. R. Brown and Taylor himself, had any previous
experience with Whirlwind I. Requesting the transfer to his group
of one or two more men possessing Whirlwind I training, Taylor
estimated that even if all effort were taken off Whirlwind I, com-
pletion of Whirlwind II would not be accomplished until January 1,
1956, two years later than scheduled. If Whirlwind I personnel
were not used, then completion of a production model would be
extended another two years .

In response to Taylor's predictions and pleas, Steve Dodd,
head of Group 64, argued that a similar lack of experienced person-
nel would interfere with and delay the planned program for the
improvement and enlargement of Whirlwind I. Whirlwind I had been
considered a "training ground for Whirlwind II," but already the
requirements for the former's program had increased "faster than
the training;" consequently, no surplus of personnel existed for.
transfer. Furthermore, he warned, dilution of "the effort on
Whirlwind I and Cape Cod might easily push Whirlwind I out to the
original time estimates for Whirlwind II." Forrester concluded

the discussion by recommending that the problem be taken under

35
consideration and investigated further by Dodd and Wieser.

12.20

The conclusions reached by Dodd and Wieser were presented at
the group leaders' meeting of June 9. Dodd, acting as spokesman,
told his colleagues that "no experienced staff could be transferred
without seriously cutting into the Cape Cod program." This could
have been particularly damaging to the over-all program, especially
if Wieser 's earlier warning — that the Cape Cod system would be

"running too late to be of any use in affecting decisions in the

37
design" of Whirlwind II — proved accurate. Forrester's decision,

given at the end of the meeting, was that a transfer of experienced

personnel from Group 64 to Group 62 would not be in "the best

interests of the laboratory." Consequently, Taylor was required to

38
train his own men and to make his assignment accordingly.

The fourth major task which faced the directors of Division VI
was the selection of a manufacturing source for the air defense com-
puter. From the middle of 1951 on, analyses were undertaken and
applications from qualified manufacturers were solicited, informally
at first. Ultimately, the Air Force placed the International
Business Machines Corporation under contract. From the outset of
contractual negotiations, IBM had preferred a prime contract. This
view was shared by MIT because the Institute's administration was

loath to expand its budget through "additional funds for large sub-

39
contracts." On the other hand, the Cambridge Research Center,

representing the Air Force in the day-to-day negotiations, believed

that a prime contract was impossible at the start because revised

Air Force regulations made difficult if not impossible the

12.21

justification of IBM as a "sole source." Consequently, John

Marchetti, speaking for the Center, recommended a temporary sub-

40
contract with Lincoln Laboratory.

Although MIT and IBM pressed the matter as a matter of prin-
ciple, using it as a means of inducing the Air Force to ease its

41
contractual procedures, both ultimately accepted a sub-contractual

relationship which, it was anticipated, would within a matter of

months be replaced by a prime contract between the Air Force and

the Corporation. Marchetti, however, made very clear the Air

Force's intent to have MIT bear primary responsibility for the

program, for it would "continue to monitor the prime contract" even

42
after the Air Force had taken over with "production money."

On October 27, 1952, the Institute issued to IBM a Letter of
Intent under the terms of which the Corporation commenced a coop-
erative project with Lincoln Laboratory. The contract was ultimately
to involve the Western Electric Corporation, the Bell Telephone
Laboratories, the Burroughs Corporation, and the System Development
Corporation, and it culminated in the development and construction
of the SAGE air defense system. The story of the success of the
Cape Cod system and the subsequent building of SAGE, however, is
not a part of the Whirlwind story. It is Whirlwind's sequel and is
a story the research and development aspects of which remained still
wrapped, a decade and a half later (at this writing), in the cloak
of classified, national security information. For these reasons,
the curtain must drop here. Project Whirlwind had run its creative,

12.22

independent course as an on-campus research and development organi-
zation and was being shunted by ever stronger external forces, such
as ONR in different ways had once sought to generate, along a course
that would bring it into closer alignment with the larger organiza-
tion of Project Lincoln and into an increasingly subordinate,
assisting role, supporting and implementing the specific techno-
logical needs of the growing military defense establishment.

It was one thing for a young graduate -student engineer to
attack an aircraft simulator problem in the closing years of a
great war, another thing to open a door onto undreamt-of, challeng-
ing, untrodden vistas of pioneering computer research and develop-
ment waiting to be traversed, and still another to become a train-
ing and development organization ten years later, trapped in the
detailed implementation work of devising third and fourth genera-
tion computers for increasingly routine, even though more complex,
military assignments. When Forrester perceived what appeared to
lie ahead of the organization he had built up, he found he was not
particularly challenged by the prospect. Firmly convinced he could
not restore the past and possessing extensive organizational experi-
ence acquired over a decade, he left the rapidly growing "computer
business" in 1956 to immerse himself in the serious, academic study
of principles and techniques of industrial and engineering organi-
zation.

12.23

Subsequently, Everett left Project Lincoln also. Convinced
the past could not be restored but. still keenly attracted to other
untrodden vistas in the realm of computer research and development,
he became instrumental in forming a new organization to probe the
engineering unknown, The MITRE Corporation. But these are other
stories, still unfolding, better told elsewhere, and not a part of
this account.

NOTES TO CHAPTER 12

1. MIT, Problems of Air Defense, Final Report, Project Charles ,
August 1, 1951, vol. I, pp. 96-122; C. R. Wieser in Lincoln
Laboratory, Quarterly Progress Report, Division 6 - Digital
Computer, June 1, 1953, pp. 6-10 — 6-11; Brief Summary of
Activity Planned by Project Lincoln During F/53, PLA-126,
Jan. 9, 1951.

2. Schedule to Contract No. Af 19 (122) -458.

3. HEADC, Operational Plan, Semiautomatic Ground Environment
System for Air Defense , March 7, 1955, p. vi.

4. Memorandum A- 124, J. W. Forrester, sub j . : "Change in
Laboratory Name," September 20, 1951.

5. J. W. Forrester, Computation Book No. 49 . p. 139, entry for
October 1, 1951.

6. Interview, J. W. Forrester and R. R. Everett by the authors,
July 31, 1963.

7. Memorandum L-32, J. W. Forrester, subj . : "Project Lincoln,
Division VI Program, July 1952- June 1953," January 7, 1952.

8. Administrative Memorandum A- 131, H. Fahnestock, subj.:
"Accounting Procedures, DIC 6886," March 26, 1952; J. W.
Forrester, Lincoln Laboratory Quarterly Progress Report ,
Division 6--Digital Computer, June 1, 1952 , pp. 6-7.

9. Memorandum 6L-173, J. W. Forrester, subj.: "Organization and
Tasks of Division 6," November 16, 1954.

10. Interview, John C. Proctor by the authors, July 13, 1964.

11. Interviews, Harris Fahnestock and John C. Proctor by the •
authors, July 15 and 13, 1964, respectively.

12. Memorandum L-32, J. W. Forrester, subj.: "Project Lincoln,
Division VI Program, July 1952-June 1953," January 5, 1952;
Memorandum L-45, J. W. Forrester and R. R. Everett, subj.:
"Condensed Summary of FY 54," June 10, 1952.

13. Memorandum L-86, C. R. Wieser, subj.: "Cape Cod System and
Demonstration," March 13, 1953.

NOTES TO CHAPTER 12 (CONTINUED)

14 o "Brief Summary of Activity Planned by Project Lincoln during
FY 53," PLA-126, January 9, 1951; Memorandum L-45, J. W.
Forrester and R. R. Everett, sub j . : "Condensed Summary of
FY 54," June 10, 1952.

15. Memorandum L-32, J. W. Forrester, sub j . : Project Lincoln,
Division VI Program, July 1952 - June 1953."

16. Memorandum M-1810, A. P. Kromer, sub j . : "Minutes of Joint-MIT-
IBM Conference, Held at Hartford, Connecticut January 20, 1953,"
January 26, 1953, p. 3.

17. Memorandum L-64, C. R. Wieser, sub j . : "ADEE Steering Committee
Meeting at Colorado Springs, October 7, 8, 1952;" Memorandum

L- 6 6, Arthur Kromer, sub j . : "Discussion of contract status
with IBM," October 17, 1952; J. W. Forrester, Computation Book
No. 53 , p. 9, entry for November 3, 1952; Memorandum L-71,
D. R. Brown, subj.: "Group Leaders' Meeting, November 24,
1952," November 24, 1952.

18. Letter, Major General D. L. Putt, Vice Commander, HEDARDC to
COMRADC, subj.: "Revision of Command Policy Pertaining to ADIS,"
May 6, 1953; letter, Lt. General Earle E. Partridge, Commander,
ARDC, to Dr. Harlan Hatcher, President, University of Michigan,
May 6, 1953.

19. Memorandum L-64, C. R. Wieser, subj.: "Meeting at Colorado
Springs, October 7, 8, 1952," October 10, 1952; Limited Memoran-
dum L-65, J. W. Forrester, October 13, 1952; Memorandum L-71,

D. R. Brown, subj.: "Group Leaders' Meeting, November 24, 1952,"
November 24, 1952.

20. Memorandum L-86, C. R. Wieser, subj.: "Cape Cod System and
Demonstration," March 13, 1953.

21. Lincoln Laboratory, Quarterly Progress Report, Division AF-24 6 —
Digital Computer, December 15, 1953, p. iii.

22. Memorandum L-129, D. R. Brown, subj.: "Group Leaders' Meeting,
December 7, 1953," December 7, 1953.

23. Project Whirlwind Summary Report No. 28. Fourth Quarter 1951,
p. 11.

24. Ibid., p. 11.

NOTES TO CHAPTER 12 (CONTINUED)

25. Memorandum L-36, R. R. Everett, subj.: "Second Bank of 1024-
Digit Storage Tubes for WWI," April 3, 1952.

26. Summary Report No. 35, Third Quarter, 1953 , pp. 5, 32-33.

27. Ibid ., p. 33.

28. Memorandum M~ 18 10, A. P. Kromer, subj.: "Minutes of Joint
MIT-IBM Conference, Held at Hartford, Connecticut January 20,
1953," January 26, 1953.

29. Memorandum M-1321, B. E. Morris, subj.: "Fourth Meeting on
Air Defense Computer," November 8, 1951, p., 4.

30. Memorandum L-45, J. W. Forrester and R..R. Everett, subj.:
"Condensed Summary of FY 54," June 10, 1952.

31. Memorandum L-30, J. W. Forrester and R. R. Everett, subj.:
"Digital Computers for Air Defense System," October 5, 1951.

32. Memorandum M-1327, D. R. Brown and B. E. Morris, subj.:
"Summary of First Four Meetings on Air-Defense Computer,"
November 9, 1951.

33. Memorandum L-32, J. W. Forrester, subj.: "Project Lincoln,
Division VI Program, July 1952 - June 1953," January 7, 1952.

34. Memorandum, R. R. Everett and H. Fahnestock to J. W. Forrester,
subj.: "Senior Staff Meeting," March 5, 1952.

35. Memorandum L-34, H. Fahnestock, subj.: "Group Leaders ' Meeting,

March 26, 1952," March 27, 1952.

36. Memorandum L-46, D. R. Brown, subj.: "Group Leaders' Meeting,
June 9, 1952," June 11, 1952.

37. Memorandum L-37, D. R. Brown, subj.: "Group Leaders' Meeting,

April 7, 1952," April 9, 1952.

38. Memorandum L-46,

39. Memorandum L-66, J. W. Forrester, subj.: "Discussion of con-
tract status with IBM," October 17, 1952; J. W. Forrester,
Computation Book No. 53 , pp. 1, 2, & 4, entries for October 17,
1952.

NOTES TO CHAPTER 12 (CONTINUED)

40. J. W. Forrester, Computation Book No. 53 , p. 7, entry for
October 27, 1952.

41. J. W. Forrester, Computation Book No. 53 , pp. 1, 2, & 4,
entries for October 17, 1952 „

42. J. W. Forrester, Computation Book No. 53 , pp. 6-7, entries
for October 22 and 27, 1952, respectively; Memorandum M-1739,
A. P. Kromer, sub j . : "Summary of IBM-MIT Collaboration,
October 27, 1952 to November 30, 1952 inclusive," December 3,
1952; Digital Computer Laboratory, MIT, Purchase Requisition,
DIC-L 33210, December 4, 1952.

Chapter Thirteen
IN RBTaOSPflCT

For all the adventures that befell Project Whirlwind and
for all the changes that occurred in the aims and procedures of
the Project, from the days of the aircraft simulator to the days
of continental air defense, it enjoyed a remarkable constancy of
identity as a team of investigators dedicated to the prosecution
of the research and development enterprise. In the continuity
of its style of carrying on its inquiries and in the depths of
its commitment to producing practical machinery, it maintained a
unity of character and a philosophy of investigation which
marked it as unique among R h. D projects.

In a sense, every H & D project is unique, of course, just
as in another sense every R&D project is representative. Cer-
tainly, Whirlwind was both, and in these respects it offers provoc-
ative clues to the nature of the historical process in general and
to the nature of the R&D process in particular. Someone has
said it is the business of the past to produce a present that is
different, and the lessons of twenty-five centuries of science
and two centuries of scientific technology suggest that further-
more it is the business of science and R & D to annihilate their
pasts to produce the novel present,

13.1

13.2

All the signs indicate this process is an evolutionary one
(marked here and there, it is true, by changes so rapid as to
seem revolutionary in their impact) , and Project Whirlwind was
no exception. Superficially the story of Whirlwind divides
quite naturally into two parts which although interdependent
and interrelated stand distinct. This distinction is the more
evident from the fact that the divisions are chronological,
covering the periods 19^ to 1950 and 19^9 to 1956. Moreover,
each period correlates with a distinct, minor era in the nation's
transition from the Second world War, through the retrenchment
of peace to the Cold War and the Korean War, and each period can
be further defined by the dependence of the Project upon either
Navy or Air Force financing and managerial assistance, interference,
and supervision.

Looking back, one sees the years between 19kk and 1950
comprising for Project Whirlwind a period of gestation, for it
was during these years that emphasis shifted from the restricted-
purpose simulator to the general-purpose computer, and it was
during these years that the computer was brought to birth as an
operating machine. Also, these years broadly delimit the period
in which the Navy played the primary role in the program, and —
perhaps of greatest importance for the climate in which H & D
operated— they approximate the interlude between tne end of the
Second World V.'ar and the beginning of the international police
action called the Korean War. This was an interlude marked by a

13.3

national policy of military retrenchment which significantly
affected the magnitude and momentum of the Project, and this has
been the period of primary concern to this study.

ihe years between 19^+9 and 1956, reflecting the anxieties
of a world divided, were marked by a re-emergence of concern
over the inadequacies of the nation's air defenses and by pro-
grams mounted to study and correct those inadequacies. One of
these programs led to the successful demonstration by Whirlwind
of the feasibility of an air-defense command and control center
equipped with the digital computer as the principal information-
coordinating element. The end result was Air Force participation
and the displacement of the Navy as the major source of funds
and purpose. Also there resulted the assimilation of the Whirl-
wind team into the Air Force program and the development of the
semi-automatic ground environment (SAGS) air defense system
which, prior to the missile age, was intended to provide max-
imum security against attack from the air.

This same concept— a centralized computer of large capacity
fed by geographically scattered radar sensors— was subsequently
modified and applied to continental defense against missile attack.
Thus the conceptions of "Command and Control" which Whirlwind had
demonstrated as feasible, and in the development of which Whirl-
wind had played a vital role, was incorporated into the national
defense structure as an essential element, further, the conceptions

13.4

of command and control were to expand well beyond military use
through application to other governmental needs and to the needs
of industry and society in general, as the computer moved in the
direction of becoming one day a true public utility which, so
proponents argued, would rank with the telephone and the water
faucet*

Such in brief was Whirlwind's national historical significance.
In addition, it was significant as an instance of the R&D process
in operation, for it was at once continuous and broadly predictable
in its research procedures, at once dis continuous and unpredictable
in its potential applications and in the inventive originality of
its engineers. The Project was at the same time highly personal
and individual with respect to the inventive and developmental tal-
ents displayed by its members in the administrative realm, the fiscal
realm, and the technical realm. It was at the same time social and
anonymous in its exploitation and coordination of the engineering
talents of its team.

While the technical story could be understood only by the
specialist, the general shape and color of this R&D enterprise
can be appreciated by the citizen, and in addressing this case
history to him the authors hope that more questions have been
raised than answered, for the rise of R & D is a recent historical
phenomenon which first began to emerge in its modern form in

13.5

eighteenth-century 3urope, and it is still too new to be well
understood. Nevertheless, with it man is already remolding the
world and moving oat into Space, and although Project Whirlwind
becomes in this perspective an obscure enterprise that most
readers will never have heard of, it presents many characteristic
features and several unusual ones the perception of which sharpens
our citizen's understanding of the R&D process and improves our
prospects of more intelligently directing it in the best interests
of our republic.

To begin, there are the accomplishments of Project Whirlwind,
and of these there are several technical accomplishments that
should be singled out. Forrester, looking back from the perspec-
tive of more than a decade, could find over a dozen devices, pro-
cesses, and applications which Whirlwind contributed or brought
to a practical working level. The details of their operation
belong in a technical engineering history of the origins, devel-
opment, and Derfection of the Semi-Automatic Ground Environment
(SAGS) air defense system, a technical story which would begin
with Whirlwind (not with the Aircraft Stability and Control An-
alyzer) and would include the next- generation ANAFS^-7 production
computer that demonstrated the worth of the Whirlwind technical
concepts in the military operation of SAGE. Although this is
not the place to describe these accomplishments in detail, they
may be enumerated briefly here to provide one essential measure

13.6

of the practical success of Project Whirlwind as an innovating
engineering enterprise.

The most famous contribution was the random-access, magnetic
core storage feature, which was to be widely employed in succeeding
generations of faster and more compact digital computers* Marginal
checking, to detect deteriorating components, was another novel and
highly practical feature. The Whirlwind computer was also first
and far ahead in its visual display facilities. One form of in-
formation output was a cathode ray tube display "capable of plot-
ting computed results on airspace maps." Associated with it was
the "light gun," or light pen, with which an operator could "write"
on the face of the cathode-ray tube display and provide new infor-
mation which the computer could store and use. As a consequence
of these two features, simultaneous man-machine interaction at will
became feasible, adding to the versatility and usefulness of the
digital computer.

Simulation techniques were perfected by which hypothetical
aircraft flights could be programmed into the computer for study
and training purposes, as a consequence the practical prospects
for digital simulation were greatly enhanced, and digital simula-
tion subsequently was richly exploited in a wide variety of fields.
The crystal matrix switch designed by David Brown, the magnetic
matrix switch developed by Kenneth 01sen,and the cryotron invented
by Dudley Buck were all Project Whirlwind products that were to see

13.7

continuing use and development later as digital computer electronics
design progressed.

The Whirlwind machine was extensively programmed to carry out
the novel procedure of self-checking, including the tasks of iden-
tifying defective components and typing out appropriate instructions
to the operator. jtSarly random tube failures posed another hurdle
that the Whirlwind project negotiated successfully, for scrutiny and
modification of tube- fabrication techniques led to dramatic in-
creases in length of tube life through procedures applicable to the
manufacture of hundreds of standard radio tube types.

The need to send radar- gathered data long distances to a
computer control center caused successful techniques to be developed
by the Lincoln Laboratory for sending digital data over telephone
lines and opened the way for later commercial applications. The
incprporation of a computer in a control network involved the im-
portant development of a practical "feedback loop," in which the
computer changed its control instructions as it received new
information and thus maintained pertinent control, as when dir-
ecting interceptor aircraft toward their targets in the early
Whirlwind air defense tests. Here lay the prophetic significance
of the L-l. and L-2 Reports of 19^7, and here was another practical
application of the computer— the feedback control loop— that would
see continuing military and commercial use in the years to come.

13.3

A related pioneering development is to be seen in the doctoral
dissertation of a Project Whirlwind engineer, William Linville, on
"sampled data control theory." Linville investigated the effects
of sampling operations upon the stability of feedback control loops

in those situations in which different users would be sharing the

2
services of a large computer to solve their separate problems.

L a st, and of profound influence upon subsequent computer
design, was the working out for Whirlwind of the intricate sys-
temic details of "synchronous parallel logic" — i.e., the trans-
mitting of electronic pulses simultaneously within the computer
feather than sequentially, while maintaining logical coherence and
control and accelerating enormously (compared to other computers of
that day) the speeds with which the computer could process its
information. As Forrester has noted, "The parallel synchronous
logic worked out for the Whirlwind computer and first appearing
in the block diagram reports done by Robert Everett and Francis
Swain set the trend for many later computer developments."

In addition to these technical accomplishments, Project
Whirlwind also demonstrated several fundamental features of the
research and development process. Two of these features of the
3 -k D process are the twin historical phenomena of "convergence"
and "divergence.* 1 There was the convergence of concerns and
enterprises involving the design of new airplanes and the design

1.5.9

of flight trainers that lured the Massachusetts Institute of Tech-
nology into preliminary involvement with the Navy, and there was
the consequent divergence represented by the findings of the aero-
dynamicists, which brought Gordon 3rown and the oervomechanisms
Laboratory into the next phase of the inquiry, as an instance
of even greater strategic import, there was the multiple conver-
gence of the various intellectual, scientific, technical, aid
mechanical traditions tnat brought the incipient computer state
of the art to the position it occupied when Crawford, fforrester,
Everett, and their associates in the Wavy and at KIT began to
explore it, aid there was the divergence characterized by the
writing of the L-l and L-2 Reports and by the abandonment of
the aircraft simulator for the computer. Still another example
of convergence was the combination of events that produced first
Valley's air defense committee and subsequently Valley's discovery
of a Whirlwind already far along in construction. The divergence
that followed appeared for a while as though it would cause
Project Whirlwind to be swallowed up by Lincoln Laboratory, but
the vitality of the former and the actions of its leaders pro-
duced another course of action.

It would be an oversimplification and a distortion of
history to assert that simple, direct cause-and-effect relationships
of a one-to-one nature exist between every convergence-divergence
sequence or between every divergence-convergence sequence of events.

13.10

But such a schema, when loosely yet carefully applied, helps to
explain the weaving of the fabric of events that typically con-
stitutes the R&D process in particular and the larger historical
process in general*

Along with the phenomena of convergence and divergence are
to be found the essential evolutionary continuities and the
equally essential revolutionary discontinuities which characterise
the purpose and direction, as well as changes in direction, of
human affairs. Fiscal, technical, aad administrative obstacles
(restriction of iffunds, erratic storage tubes, inspection visits
to the Project) tested continuity of purpose and validity of
policy judgments at the same time that they provoked inquiry into
alternative directions and courses of action.

None of these remarks is intended to suggest that the 3 & D
process, or for that matter history itself, is fundamentally pre-
dictable or determinant in character, but they do indicate the
availability of analytical tools which render the conduct of
research and development more understandable than national policy,
for example, hitherto has recognized it to be. It is possible to
follow the history of Project whirlwind while laboring under the
traditional misunderstanding of how science and technology are
thought to interact, but more insightful avenues lie open before
us. Among these is the perception that the measure of R & D —
its proper operation — can not be taken by applying either the

13.11

traditional , impractical standards of "Pure Science" or the tradi-
tional, practical standards of technology.

Some of the difficulties the mathematicians and the physics-
oriented scientists of the late Forties encountered when trying
to evaluate Project Whirlwind as an R & D enterprise arose from
their commitment to the historic values of pure science. While
these were appropriate enough for science, they were not appro-
priate for R & D. Thus, when the question was asked whether
Whirlwind might not be poor biscuits because it was trying to be
cake, the possibility was not seriously considered that engin-
eering, instead of science, might be the cake. The curious
notion has long prevailed that products of the mind alone are
somehow loftier and mightier than the products of mind and hand
combined. While the "biscuit-cake" analogy (perhaps 'metaphor'
would serve better) was not pushed so far as to fault Forrester
for not being committed to a Newton-like or Sinstein-like intel-
lectual enterprise, the conception that ma th-and- physics
standards might not be applicable at all apparently did not
occur to the investigators.

Similarly, the distinctions drawn between the Project
Whirlwind engineers 1 ways of proceeding and the Institute for
Advanced Study scientists' ways of proceeding apparently fell
on deaf ears, for all the difference it made in the way ONR
or the Lincoln Laboratory conducted their affairs. (One is

13.12

tempted here to speculate on how the Lincoln Laboratory would
have conducted its affairs had Forrester ever taken the helm.)
It is appropriate to ask how much more attention might well have
been paid to those distinctions had policy-makers felt more keenly
the importance of distinguishing more perceptively between the mix
of goals and procedures subscribed to by the Whirlwind leadership
and the mix of goals and procedures embraced by the builders of the
IAS computer. At issue here is the difference between basic re-
search of the pure science tradition and that which is sometimes
called "developmental research" in the R & D tradition.

The following archetypal metaphor between the basic researcher
and the developmental researcher may make it more clear why the
Whirlwind engineers' expensive way of proceeding was so difficult
for QWR and its consultants to appraise. (In all fairness, it
should be added that Air iforce endorsement was born not out of
any depth of understanding, for that of the Air Force was no
deeper than the Navy's, but out of a desperate practical need.)

The Basic Researcher may work with pencil and paper and
theoretical equations describing the ideali z ed phenomena. He
may 3 lso work with laboratory equipment which produces a desired,
artificial, controlled environment in which certain phenomena are
isolated and manipulated for closer study. The basic researcher
is interested primarily in understanding and explaining the

15.13

phenomena he is examining in that environment, and he may achieve
this understanding by proceeding from his paperwork to his lab-
oratory equipment, from his laboratory equipment to his ex-
perimental results, and from his experimental results to further
paperwork describing the test results, analyzing them, and re-
lating them to existing and emerging understanding.

The Developmental Researcher, on the other hand, is more
interested in devising hardware or perfecting a process and
testing how well it performs. He is interested in applying
the research understanding already gleaned, together with en-
gineering know-how, to the problem of making hardware that will
work. To this end, he may subject the hardware to the artificial,
controlled esvironment that isolates and identifies the roles the
phenomena play* Interested in what happens to his equipment when
it is subjected to the environmental stresses imposed by "real"
working conditions, he develops hardware that will exploit and
profit from his scientific understanding of the phenomena a nd
from his engineering knowledge of the materials and the working
conditions. Often, in the course of subjecting his hardware to
a working environment, he uncovers new phenomena or new roles
that known phenomena are seen to play, thereby creating new
paperwork problems and opening up new lines of inquiry for the
basic researcher to pursue. Likewise, the basic researcher
while concentrating on understanding the phenomena may turn up

13.3A

new information and vital questions of immediate or delayed use
to the developmental researcher.

In this paradigm of the basic researcher and the devel-
opmental researcher it has been assumed, to make the exposition
easier, that these are two different individuals. However, it
is possible for the same individual to play both roles, and the
more sophisticated the .R&D problem (such as that of devel-
oping the magnetic cores, for example), the more likely the
same individual or the same group of investigators will play
both roles before they are finished. The extraordinary insis-
tence with which Forrester and Sverett maintained their policy
of circulating information heavily and rapidly and soon, in
wfcitten form, among the members of the Project tended to en-
courage the interplay of roles, just as did the availability
of quality materials and services. Although theirs was nom-
inally an engineering enterprise, Everett when working out
his historically influential block diagrams or Forrester,
and subsequently Papian, when working with the phenomena of
magnetic remanence, were deeply involved in both basic and
developmental research because they were unwilling to stop
after the conception had been set down .

The ratio of supporting-services costs to technical costs
was high in Project Whirlwind, md it exposed the Projectto ac-
cusations of "gold plating" practices by those who held to
a philosophy dominated by conceptions centering around an

13.15

unexamined commitment to what might be called an "economy of
scarcity." To such critics, the only alternative was an "economy
of plenty," which could really be nothing more than an irresponsible
fool's paradise inasmuch as all costs had to be recovered, and dir-
ectly, even if it meant robbing Peter to pay Paul. Indeed, that
was precisely what Project Whirlwind was flagrantly doing, with its
insatiable demands for more money and more money and more money!
3ut from Forrester's and Everett's point of view, one need
not be trapped into choosing only between these two polar opposites.
"There is a need to subordinate these problems of balance," 3verett
remarked, "to a philosophy of creative force and inherent growth
which tells you how to proportion your services to your technical
effort." The whirlwind project demonstrated this, as far as
Forrester and Everett were concerned in after years, yet the
lesson remained unappreciated in the thronging halls of R 'u D,
and this, too, the two men felt keenly. Although a senior executive
from the International Business Machines Corporation had asked,
after a first visit to the v/hirlwind installation, "How do you
achieve so much with so little?" it was as clear in the early Fifties
when he visited the Barta building as it was a decade and a half
later that the essential intangibles of the Whirlwind philosophy
of conducting research and development had gone unappreciated.
That philosophy involved far more than a services costs to technical
costs ratio, as this case history has shown.

13.16

In a world without pain, where value judgments did not compete for
selection and where the proper course of action was always discern-
ible, there would be no need for the harassment of inspection
visits by third parties, but in the world of Hi D where Project
Whirlwind dwelt such inspections play a healthy, tonic role,
provided there is basic policy agreement about how research and
development affairs should be conducted. To cite am obvious
failure to agree on basic policy, an inspection team which crit-
icized a fundamental research laboratory for its failure to main-
tain engineering development projects on a massive scale would
only be throwing sand in the gears, however well intent ioned and
sincere its motives. Similarly, a legislator who criticized
the National Science foundation, for example, for supporting im-
practical, "out in the wild blue yonder" basic research with
hard-earned public zax dollars either would be making what he
judged to be the right "political sounds" for the home folks in
an election year or else would be demonstrating that he really
did not share the basic policy attitudes which technically
versed administrators have fo md to be viable over the many
centuries that science has flourished.

No such gross mistakes in selecting inspection visitors
appear to have occurred in v.'hirlwind's case, for qualified per-
sonnel were selected, for the most part. Unfortunately, the
qualified person is apt to be a specialist, and the specialist
is as much a custodian of essential knowledge in his field as
he is a creative entrepreneur. At stake here is the philosophical

13.17

commitment which science and engineering long ago made when they
developed the habit of seeking the judgment of professional peers
when assessing the worth of an enterprise. Whirlwind's experience
with inspection visitors does not contradict the historical gener-
alization that the peer system has proved inherently cautious and
conservative, preferring only modest departures and limited leaps
into the future.

Project Whirlwind appears to have been sufficiently the vic-
tim of this professional timidity syndrome to run into trouble,
but the true heart of the trouble lay neither with the irroject's
unorthodox modes of conducting its R&D affairs nor with the
premium placed upon conservative judgment by the inspection pro-
cess itself, --father, it lay in the fact that, for all their
sophisticated abilities to innovate— indeed, innovation is their
reason for being— the basic and the applied sciences have been
unable to develop reliable techniques for distinguishing the
seer from the fool. Consequently, they play it safe and endorse
the modest change which does not break sharply with tradition.
It w 9 s just thjis sort of crippling inability which caused the
Ad Hoc Panel to criticise Whirlwind for its lack of a mission
on the one hand while complaining that not enough attention was
being given by computer projects to "real time' 1 applications
on the other.

Nevertheless, appraisal by peers is likely to remain in
force, and in the R&D area, where the prospect of practical

13.18

accomplishments is a guiding consideration, it is probably a helpful
technique and provides a useful mode of healthy external criticism.
Because Project Whirlwind succeeded does not mean that it was
inevitably destined to succeed. It was, like all challenges, a
creature of human endeavor. It did achieve its goal, however, and
it did accelerate computer progress both by the concepts it demon-
strated and by the talented engineers it developed. As a conse-
quence, hindsight permits one to hail it as a model of R & D. But
had its funds been cut off, it would likely have joined that ever
growing number of military R & D enterprises which come to an end
on the scrap heap. Whatever the Project's vitality (and it was
considerable), and whatever the resourcefulness of its leadership,
Whirlwind was in part master of its fate and in part a creature of
larger circumstance. The words of another observer of American
science come to mind. Although uttered in another context, they
are relevant to /Whirlwind. "We must realize," remarked Vtf. Carey
of the Bureau of the Budget in 1957 » "that when science and
education become instruments of public pplicy," — and, we should
add, of public funding — "pledging their fortunes to it, an un-
stable equilibrium is established. Public policy is, almost by
definition, the most transient of phenomena, subject from be-
ginning to end to the vagaries of political dynamism. The
budget of a government, under the democratic process, is an
expression of the objectives, aspirations, and social values of
a people in a given web of circumstances. To claim stability

13.19

for such a product is to claim too much* In such a setting,

5
science and education become soldiers of fortune*" The story

of Project Vftiirlwind, as well as the story of what became of

this R&D enterprise in the years after 1956, is that of a

soldier of fortune.

NOTSS TO CHAFER 13

1. Lt», J. W. Forrester to C. W. Farr, December 18, 1967*

2* For an introduction into the possibilities of using the
time-shared computer, the reader is directed to: D. F.
Parkhill, The Challenge of the Computer Utility « (Addison-
riesley, 196*6) •

3. Ltr., J. tf. Forrester to C. W. Farr, December 18, 19&7«

4. Interview, 3. R. Bverett by the authors, October 26, 1967.

5. Scientific Manpower — » 1957 * National Science Foundation
Document No. NSF-53-21, p. 25*

MIT

CORPORATE

BOX 2'

BEDFOF

MASSACHUSETTS 017'

<EX><gT^2^1>glXEx><EI X^XXlIaLXaLIauXaI^